โŒ

Normal view

There are new articles available, click to refresh the page.
Today โ€” 17 December 2025Main stream

Hyper Foundation proposes burning all HYPE in its Hyperliquid Assistance Fund

16 December 2025 at 23:21
The Hyper Foundation has proposed treating all HYPE held in its Hyperliquid Assistance Fund as permanently burned, removing the tokens from supply via validator vote. A governance proposal released by the Hyper Foundation would exclude all HYPE held in theโ€ฆ

Yesterday โ€” 16 December 2025Main stream

As HHS restricts telework, CDC asks employees to โ€˜bypassโ€™ reasonable accommodation process

The Centers for Disease Control and Prevention is exploring workarounds to a new Department of Health and Human Services policy, which sets stricter rules on telework as a reasonable accommodation for employees with disabilities.

The HHS policy states all requests for telework, remote work, or reassignment must be reviewed and approved by an assistant secretary or a higher-level official โ€” a decision that is likely to slow the approval process.

The new policy, as Federal News Networkย reported last week, generally restricts employees from using telework as an โ€œinterim accommodation,โ€ while the agency processes their reasonable accommodation request. But faced with a backlog of more than 3,000 reasonable accommodation cases, the CDC is taking an ad-hoc approach to granting temporary medical telework.

According to four CDC employees, supervisors have instructed staff to email their medical documentation directly to Lynda Chapman, the agencyโ€™s chief operating officer, to โ€œbypassโ€ the traditional reasonable accommodation system, and receive up to 30 days of telework as an interim accommodation.

โ€œThe instructions are for you to email her a letter from your doctor, and she will be the judge of if you can telework for up to 30 days,โ€ a CDC employee told Federal News Network.

A second CDC employee shared a photo with Federal News Network showing that Chapman was recently added as an authorized official to review the employeeโ€™s reasonable accommodation materials.

โ€œI think this is very problematic, because she really should not be evaluating peopleโ€™s health needs,โ€ the employee said. โ€œIt should be an RA specialist.โ€

Two CDC employees told Federal News Network that Chapman is only approving interim telework in a few circumstances โ€” including recovery from surgery, pregnancy or chemotherapy.

This week, the CDC hosted a series of โ€œoffice hoursโ€ sessions with supervisors. During these question-and-answer sessions, the agencyโ€™s Office of Human Resources gave supervisors more information about the new reasonable accommodation process.

โ€œI was requested to share my medical information via personal email to Lynda Chapman,โ€ a CDC employee wrote in a screengrab of one of these Q&A sessions. โ€œWhen I questioned her role prior to sending my file, she denied my request.โ€

Linnet Griffiths, a former senior advisor to CDCโ€™s chief operating officer who left the agency in April, told Federal News Network that the agency had a โ€œrobust systemโ€ for processing reasonable accommodations, but said many employees who carried out this work were targeted by reductions in force.

โ€œThey have gotten rid of all the RA staff and EEO offices, which is extremely disturbing, because the department already had a lot of RA cases, a backlog that was unbelievable. They were understaffed,โ€ Griffiths said.

Griffiths said human resources employees who left CDC and HHS received specialized training to ensure the agency was following โ€œdue diligenceโ€ when processing reasonable accommodations. That included making sure the agency complied with requirements under the Americans with Disabilities Act and the Rehabilitation Act.

โ€œWe had in-house doctors that were specifically trained to review that information and approve whether someone should be eligible for full-time telework,โ€ Griffiths said. โ€œThe [CDC] chief operating officer or someone that does not have medical expertise reviewing that information was something we would never do.โ€

Griffiths said that reasonable accommodation requests often took months to get approved during her tenure at CDC, but said the agency granted full-time telework as an interim accommodation, when deemed necessary by an employeeโ€™s doctor.

In a readout from one of these โ€œoffice hoursโ€ meetings, obtained by Federal News Network, agency leadership told supervisors that, under the new HHS policy, they cannot approve interim telework requests, even in cases where telework has already been identified โ€œas the only effective accommodation.โ€

According to the readout, supervisors were told that expiring telework agreements, granted as an interim accommodation, will not be renewed, and that employees will have to either return to the office or use leave. Supervisors were told that these outcomes were โ€œnot denials.โ€ Instead, supervisors were instructed to tell employees that telework could not be granted.

โ€œSeveral supervisors noted this functions like a denial in practice,โ€ the readout states. โ€œLeadership acknowledged legal risks associated with forcing or effectively compelling leave.โ€

During one of the sessions, multiple supervisors raised concerns that agency guidance conflicts with federal disability law.

โ€œLeadership directed supervisors to stop discussing legal issues during the session, stating it was not the forum for legal discussion, and advised that such concerns should be raised offline through supervisory channels,โ€ the readout states.

According to the readout, when supervisors raised concern about personal legal exposure, leadership advised supervisors to โ€œconsider obtaining professional liability insurance.โ€

โ€œOverall, the call left many supervisors concerned about how to lawfully provide interim accommodations, the lack of written guidance, and how to avoid harm to employees while complying with the direction given,โ€ the readout states.

A CDC spokesperson said in a statement that interim accommodations, including telework, โ€œmay be provided while cases move through the reasonable-accommodation process toward a final determination.โ€

โ€œThis has always been the case,โ€ the CDC spokesperson said.

โ€˜Outsized harmโ€™

Several Senate Democrats, in a letter led by Sens. Tim Kaine (D-Va.) and Raphael Warnock (D-Ga.), said the new HHS policy โ€œwill inflict outsized harm on workers with disabilities,โ€ including employees with chronic diseases and compromised immune systems.

โ€œThe federal government is a major employer of people with disabilities. It is bound by law not to discriminate against those workers and to take steps to increase employment of workers with disabilities,โ€ the senators wrote.

The senators said department employees โ€œhave been harmedโ€ by its new reasonable accommodation policy.

The senators wrote that an HHS employeeโ€™s telework accommodation because of a high-risk pregnancy was rescinded. On the day she was supposed to report back to the office, she was rushed to the emergency room by ambulance.

Kaine and Warnock said a disabled veteranโ€™s post-traumatic stress disorder (PTSD) was exacerbated by a shooting at the CDCโ€™s headquarters this summer, but their telework accommodation was denied, approved, then denied again, โ€œleaving them without direction or guidance.โ€

HHS Press Secretary Emily Hilliard said in a statement that the department โ€œwill respond directly to the senators.โ€

โ€œInterim accommodations, like telework, may be provided while cases move through the reasonable-accommodation process toward a final determination. The Department remains committed to processing these requests as quickly as possible,โ€ Hilliard said.

The post As HHS restricts telework, CDC asks employees to โ€˜bypassโ€™ reasonable accommodation process first appeared on Federal News Network.

ยฉ AP Photo/David Goldman

Veza Extends Reach to Secure and Govern AI Agents

16 December 2025 at 14:06

Veza has added a platform to its portfolio that is specifically designed to secure and govern artificial intelligence (AI) agents that might soon be strewn across the enterprise. Currently in the process of being acquired by ServiceNow, the platform is based on an Access Graph the company previously developed to provide cybersecurity teams with a..

The post Veza Extends Reach to Secure and Govern AI Agents appeared first on Security Boulevard.

Why the CIO is becoming the chief autonomy officer

16 December 2025 at 13:14

Last quarter, during a board review, one of our directors asked a question I did not have a ready answer for. She said, โ€œIf an AI-driven system takes an action that impacts compliance or revenue, who is accountable: the engineer, the vendor or you?โ€

The room went quiet for a few seconds. Then all eyes turned toward me.

I have managed budgets, outages and transformation programs for years, but this question felt different. It was not about uptime or cost. It was about authority. The systems we deploy today can identify issues, propose fixes and sometimes execute them automatically. What the board was really asking was simple: When software acts on its own, whose decision is it?

That moment stayed with me because it exposed something many technology leaders are now feeling. Automation has matured beyond efficiency. It now touches governance, trust and ethics. Our tools can resolve incidents faster than we can hold a meeting about them, yet our accountability models have not kept pace.

I have come to believe that this is redefining the CIOโ€™s role. We are becoming, in practice if not in title, the chief autonomy officer, responsible for how human and machine judgment operate together inside the enterprise.

Even the recent research from Boston Consulting Group notes that CIOs are increasingly being measured not by uptime or cost savings but by their ability to orchestrate AI-driven value creation across business functions. That shift demands a deeper architectural mindset, one that balances innovation speed with governance and trust.

How autonomy enters the enterprise quietly

Autonomy rarely begins as a strategy. It arrives quietly, disguised as optimization.

A script closes routine tickets. A workflow restarts a service after three failed checks. A monitoring rule rebalances traffic without asking. Each improvement looks harmless on its own. Together, they form systems that act independently.

When I review automation proposals, few ever use the word autonomy. Engineers frame them as reliability or efficiency upgrades. The goal is to reduce manual effort. The assumption is that oversight can be added later if needed. It rarely is. Once a process runs smoothly, human review fades.

Many organizations underestimate how quickly these optimizations evolve into independent systems. As McKinsey recently observed, CIOs often find themselves caught between experimentation and scale, where early automation pilots quietly mature into self-operating processes without clear governance in place.

This pattern is common across industries. Colleagues in banking, health care and manufacturing describe the same evolution: small gains turning into independent behavior. One CIO told me their compliance team discovered that a classification bot had modified thousands of access controls without review. The bot had performed as designed, but the policy language around it had never been updated.

The issue is not capability. It is governance. Traditional IT models separate who requests, who approves, who executes and who audits. Autonomy compresses those layers. The engineer who writes the logic effectively embeds policy inside code. When the system learns from outcomes, its behavior can drift beyond human visibility.

To keep control visible, my team began documenting every automated workflow as if it were an employee. We record what it can do, under what conditions and who is accountable for results. It sounds simple, but it forces clarity. When engineers know they will be listed as the manager of a workflow, they think carefully about boundaries.

Autonomy grows quietly, but once it takes root, leadership must decide whether to formalize it or be surprised by it.

Where accountability gaps appear

When silence replaces ownership

The first signs of weak autonomy are subtle. A system closes a ticket and no one knows who approved it. A change propagates successfully, yet no one remembers writing the rule. Everything works, but the explanation disappears.

When logs replace memory

I saw this during an internal review. A configuration adjustment improved performance across environments, but the log entry said only executed by system. No author, no context, no intent. Technically correct, operationally hollow.

Those moments taught me that accountability is about preserving meaning, not just preventing error. Automation shortens the gap between design and action. The person who creates the workflow defines behavior that may persist for years. Once deployed, the logic acts as a living policy.

When policy no longer fits reality

Most IT policies still assume human checkpoints. Requests, approvals, hand-offs. Autonomy removes those pauses. The verbs in our procedures no longer match how work gets done. Teams adapt informally, creating human-AI collaboration without naming it and responsibility drifts.

There is also a people cost. When systems begin acting autonomously, teams want to know whether they are being replaced or whether they remain accountable for results they did not personally touch. If you do not answer that early, you get quiet resistance. When you clarify that authority remains shared and that the system extends human judgment rather than replaces it โ€” adoption improves instead of stalling.

Making collaboration explicit

To regain visibility, we began labeling every critical workflow by mode of operation:

  • Human-led โ€” people decide, AI assists.
  • AI-led โ€” AI acts, people audit.
  • Co-managed โ€” both learn and adjust together.

This small taxonomy changed how we thought about accountability. It moved the discussion from โ€œwho pressed the button?โ€ to โ€œhow we decided together.โ€ Autonomy becomes safer when human participation is defined by design, not restored after the fact.

How to build guardrails before scale

Designing shared control between humans and AI needs more than caution. It requires architecture. The objective is not to slow automation, but to protect its license to operate.

Define levels of interaction

We classify every autonomous workflow by the degree of human participation it requires:

  • Level 1 โ€“ Observation: AI provides insights, humans act.
  • Level 2 โ€“ Collaboration: AI suggests actions, humans confirm.
  • Level 3 โ€“ Delegation: AI executes within defined boundaries, humans review outcomes.

These levels form our trust ladder. As a system proves consistency, it can move upward. The framework replaces intuition with measurable progression and prevents legal or audit reviews from halting rollouts later.

Create a review council for accountability

We established a small council drawn from engineering, risk and compliance. Its role is to approve accountability before deployment, not technology itself. For every level 2 or level 3 workflow, the group confirms three things: who owns the outcome, what rollback exists and how explainability will be achieved. This step protects our ability to move fast without being frozen by oversight after launch.

Build explainability into the system

Each autonomous workflow must record what triggered its action, what rule it followed and what threshold it crossed. This is not just good engineering hygiene. In regulated environments, someone will eventually ask why a system acted at a specific time. If you cannot answer in plain language, that autonomy will be paused. Traceability is what keeps autonomy allowed.

Over time, these practices have reshaped how our teams think. We treat autonomy as a partnership, not a replacement. Humans provide context and ethics. AI provides speed and precision. Both are accountable to each other.

In our organization we call this a human plus AI model. Every workflow declares whether it is human-led, AI-led or co-managed. That single line of ownership removes hesitation and confusion.

Autonomy is no longer a technical milestone. It is an organizational maturity test. It shows how clearly an enterprise can define trust.

The CIOโ€™s new mandate

I believe this is what the CIOโ€™s job is turning into. We are no longer just guardians of infrastructure. We are architects of shared intelligence defining how human reasoning and artificial reasoning coexist responsibly.

Autonomy is not about removing humans from the loop. It is about designing the loop on how humans and AI systems trust, verify and learn from each other. That design responsibility now sits squarely with the CIO.

That is what it means to become the chief autonomy officer.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?

Post-Quantum Cryptography (PQC): Application Security Migration Guide

16 December 2025 at 06:00

The coming shift to Post-Quantum Cryptography (PQC) is not a distant, abstract threatโ€”it is the single largest, most complex cryptographic migration in the history of cybersecurity. Major breakthroughs are being made with the technology. Google announced on October 22nd, โ€œresearch that shows, for the first time in history, that a quantum computer can successfully run a verifiable algorithm on hardware, surpassing even the fastest classical supercomputers (13,000x faster).โ€ It has the potential to disrupt every industry. Organizations must be ready to prepare now or pay later.ย 

The post Post-Quantum Cryptography (PQC): Application Security Migration Guide appeared first on Security Boulevard.

AI ์„ฑ๊ณต์˜ ์ถœ๋ฐœ์ , ๊ธฐ์—… ๋ฐ์ดํ„ฐ ์ „๋žต์„ ๋ฐ”๊พธ๋Š” 8๊ฐ€์ง€ ์›์น™

16 December 2025 at 02:56

IBM์˜ ๋ฐ์ดํ„ฐ ๋‹ด๋‹น ๋ถ€์‚ฌ์žฅ ๊ฒธ ์ตœ๊ณ ๋ฐ์ดํ„ฐ์ฑ…์ž„์ž(CDO)์ธ ์—๋“œ ๋Ÿฌ๋ธ”๋ฆฌ๋Š” ํ˜์‹ ์ ์ธ AI ์ „๋žต์„ ๊ตฌ์ถ•ํ•˜๋ ค๋Š” ์กฐ์ง์ด๋ผ๋ฉด ๋ฌด์—‡๋ณด๋‹ค ๊ฒฝ์Ÿ๋ ฅ ์žˆ๋Š” ๋ฐ์ดํ„ฐ ์ „๋žต์„ ๋จผ์ € ๊ฐ–์ถฐ์•ผ ํ•œ๋‹ค๊ณ  ์กฐ์–ธํ•œ๋‹ค. AI๋ฅผ ํ™•์žฅํ•˜๋Š” ๊ด€์ ์—์„œ ๋ณด๋ฉด ๋ฐ์ดํ„ฐ๋Š” ๊ทผ๊ฐ„์ด๊ธฐ ๋•Œ๋ฌธ์ด๋‹ค.

ํ•˜์ง€๋งŒ ๋Ÿฌ๋ธ”๋ฆฌ์— ๋”ฐ๋ฅด๋ฉด, ์‹ค์ œ๋กœ AI ๋ชฉํ‘œ์— ๋ถ€ํ•ฉํ•˜๋Š” ๋ฐ์ดํ„ฐ ์•„ํ‚คํ…์ฒ˜๋ฅผ ๊ฐ–์ถ˜ ์กฐ์ง์€ ๋งŽ์ง€ ์•Š๋‹ค. ๋Œ€์‹  ์ผ๊ด€๋œ ๋ฐ์ดํ„ฐ ํ‘œ์ค€์— ๋”ฐ๋ผ ๊ด€๋ฆฌ๋˜์ง€ ์•Š๋Š” ์‚ฌ์ผ๋กœํ˜• ๋ฐ์ดํ„ฐ๊ฐ€ ์—ฌ์ „ํžˆ ์กด์žฌํ•œ๋‹ค. ์ด๋Š” ์ „์‚ฌ์  ์ธ๊ณต์ง€๋Šฅ ๋ฐฐํฌ๋ฅผ ์ง€์›ํ•˜๊ธฐ๋ณด๋‹ค๋Š” ํŠน์ • ์‹œ์ ์˜ ์˜์‚ฌ๊ฒฐ์ •์„ ์ œ๊ณตํ•˜๊ธฐ ์œ„ํ•ด ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ๋‹จ์œ„๋กœ IT ํ™˜๊ฒฝ์„ ๊ตฌ์ถ•ํ•ด ์˜จ ๊ธฐ์กด ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ๋ฐ์ดํ„ฐ ์ „๋žต์˜ ๊ฒฐ๊ณผ๋‹ค.

IBM์ด 2025๋…„์— ๋ฐœํ‘œํ•œ ๋ณด๊ณ ์„œ๋Š” ๋งŽ์€ ์กฐ์ง์ด ๋ฐ์ดํ„ฐ ๋ฌธ์ œ๋กœ ์–ด๋ ค์›€์„ ๊ฒช๊ณ  ์žˆ์Œ์„ ๋ณด์—ฌ์ค€๋‹ค. ์ด ์กฐ์‚ฌ์— ๋”ฐ๋ฅด๋ฉด ์ „ ์„ธ๊ณ„ 1,700๋ช…์˜ CDO ๊ฐ€์šด๋ฐ, ๋ฐ์ดํ„ฐ๊ฐ€ ์ƒˆ๋กœ์šด AI ๊ธฐ๋ฐ˜ ์ˆ˜์ต์›์„ ์ง€์›ํ•  ์ˆ˜ ์žˆ๋‹ค๊ณ  ํ™•์‹ ํ•œ ์‘๋‹ต์ž๋Š” 26%์— ๊ทธ์ณค๋‹ค.

๋Ÿฌ๋ธ”๋ฆฌ๋Š” ํ•ด๋ฒ•์œผ๋กœ ๋ฐ์ดํ„ฐ๊ฐ€ ์ƒ์„ฑ๋˜๋Š” ์œ„์น˜์™€ ๊ด€๊ณ„์—†์ด ๋™์ผํ•œ ํ‘œ์ค€๊ณผ ๊ฑฐ๋ฒ„๋„Œ์Šค, ๋ฉ”ํƒ€๋ฐ์ดํ„ฐ๋ฅผ ์ ์šฉํ•˜๋Š” ํ†ตํ•ฉ ์ „์‚ฌ ๋ฐ์ดํ„ฐ ์•„ํ‚คํ…์ฒ˜๊ฐ€ ํ•„์š”ํ•˜๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ฐ์ดํ„ฐ ์ „๋žต์„ ์—…๋ฐ์ดํŠธํ•ด์•ผ ํ•œ๋‹ค๋Š” ์ธ์‹์€ ๋Ÿฌ๋ธ”๋ฆฌ๋งŒ์˜ ๊ฒฌํ•ด๊ฐ€ ์•„๋‹ˆ๋‹ค. IDC ๊ธ€๋กœ๋ฒŒ ๋ฐ์ดํ„ฐ์Šคํ”ผ์–ด ๋ฐ ๊ธ€๋กœ๋ฒŒ ์Šคํ† ๋ฆฌ์ง€์Šคํ”ผ์–ด ์—ฐ๊ตฌ ํ”„๋กœ๊ทธ๋žจ์˜ ์—ฐ๊ตฌ ๋งค๋‹ˆ์ €์ด์ž 2025๋…„ ๋ณด๊ณ ์„œ โ€˜์ƒ์„ฑํ˜• AI ์‹œ๋Œ€์˜ ์ฝ˜ํ…์ธ  ์ƒ์„ฑโ€™ ๊ณต๋™ ์ €์ž์ธ ์• ๋ค ๋ผ์ดํŠธ๋Š” โ€œAI๋Š” ๋ฐ์ดํ„ฐ์˜ ํ™œ์šฉ ๋ฐฉ์‹๋ฟ ์•„๋‹ˆ๋ผ ํ™œ์šฉ ๋ชฉ์ ๊ณผ ๊ฐ€์น˜๊ฐ€ ์ฐฝ์ถœ๋˜๋Š” ์ง€์ ๊นŒ์ง€ ๋ฐ”๊พธ๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ๋Œ€๋ถ€๋ถ„์˜ ์กฐ์ง์€ ์ด๋Ÿฌํ•œ ๋ณ€ํ™”์— ๋งž์ถฐ ๋ฐ์ดํ„ฐ ์ „๋žต์„ ํ˜„๋Œ€ํ™”ํ•  ํ•„์š”๊ฐ€ ์žˆ๋‹คโ€๋ผ๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

๋ผ์ดํŠธ๋Š” โ€œ๊ธฐ์กด ๋ฐ์ดํ„ฐ ์ „๋žต์€ ๋ณด๊ณ , BI, ์ž๋™ํ™”๋ฅผ ์œ„ํ•ด ์„ค๊ณ„๋์ง€๋งŒ, AI๋Š” ๋ฐ˜๋ณต์ ์ด๊ณ  ๋ชจ๋ธ ์ค‘์‹ฌ์ ์ธ ์›Œํฌํ”Œ๋กœ๋ฅผ ๊ตฌ๋™ํ•  ์ˆ˜ ์žˆ๋Š” ํ›จ์”ฌ ๋” ์—ญ๋™์ ์ด๊ณ  ์„ธ๋ถ„ํ™”๋œ ์‹ค์‹œ๊ฐ„ ๋ฐ์ดํ„ฐ ํŒŒ์ดํ”„๋ผ์ธ์„ ์š”๊ตฌํ•œ๋‹คโ€๋ผ๋ฉฐ โ€œ์ด๋Š” ์ •์ ์ธ ๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค์—์„œ ๋ฒ—์–ด๋‚˜ ์ง€์†์ ์ธ ๋ฐ์ดํ„ฐ ํ’ˆ์งˆ ๋ชจ๋‹ˆํ„ฐ๋ง, ๊ฐ•ํ™”๋œ ๋ฉ”ํƒ€๋ฐ์ดํ„ฐ์™€ ๊ณ„๋ณด ์ถ”์ , ๊ทธ๋ฆฌ๊ณ  ์ผ์‹œ์ ยท์บ์‹œยท์ €์žฅ ๋ฐ์ดํ„ฐ๊ฐ€ ํ˜ผ์žฌ๋œ AI ํŠน์„ฑ์„ ๋ฐ˜์˜ํ•œ ๋ณด์กด ์ •์ฑ…์œผ๋กœ์˜ ์ „ํ™˜์„ ์˜๋ฏธํ•œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์ด์–ด ๊ทธ๋Š” โ€œAI ์‹œ๋Œ€์—๋Š” ๋ชจ๋“  ๋ฐ์ดํ„ฐ๋ฅผ ์ˆ˜์ง‘ํ•˜๊ณ  ์ €์žฅํ•˜๋ ค๋Š” ์‚ฌ๊ณ ๋ฐฉ์‹์—์„œ ๋ฒ—์–ด๋‚˜, ๋น„์šฉ๊ณผ ์œ„ํ—˜, ๊ทธ๋ฆฌ๊ณ  ๋‹ฌ์„ฑํ•˜๋ ค๋Š” ๊ตฌ์ฒด์ ์ธ AI ์„ฑ๊ณผ๋ฅผ ๊ท ํ˜• ์žˆ๊ฒŒ ๊ณ ๋ คํ•˜๋Š” ๋ชฉ์  ์ง€ํ–ฅ์  ๋ฐ์ดํ„ฐ ์ „๋žต์œผ๋กœ ์ง„ํ™”ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋ง๋ถ™์˜€๋‹ค.

์„ฑ์ˆ™๋„๊ฐ€ ๋†’์€ ๋ฐ์ดํ„ฐ ๊ธฐ๋ฐ˜

๋Œ€๋ถ€๋ถ„์˜ ์กฐ์ง์€ ์•„์ง ์ด๋Ÿฌํ•œ ๋ชฉํ‘œ์™€ ๊ฑฐ๋ฆฌ๊ฐ€ ๋ฉ€๋‹ค. IDC์˜ ์• ๋ค ๋ผ์ดํŠธ๋Š” โ€œ๋งŽ์€ ์กฐ์ง์ด AI ํ™œ์šฉ ์‚ฌ๋ก€๋ฅผ ์ง€์›ํ•˜๋Š” ๋ฐ ํ•„์š”ํ•œ โ€˜์ ์ ˆํ•œโ€™ ๋ฐ์ดํ„ฐ๋ฅผ ํ™•๋ณดํ•˜๋Š” ๋ฐ ์—ฌ์ „ํžˆ ์–ด๋ ค์›€์„ ๊ฒช๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ์ด๋Š” ๋ฐ์ดํ„ฐ์˜ ์ถฉ๋ถ„ํ•œ ์–‘, ์ ์ ˆํ•œ ํ’ˆ์งˆ, ๋˜๋Š” ํ•„์š”ํ•œ ๋งฅ๋ฝ ๋ฉ”ํƒ€๋ฐ์ดํ„ฐ ์ค‘ ์–ด๋А ํ•˜๋‚˜๋ผ๋„ ๋ถ€์กฑํ•œ ๊ฒฝ์šฐ๋ฅผ ํฌํ•จํ•œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ทธ๋Š” ์ด์–ด โ€œIDC ์—ฐ๊ตฌ์™€ ์—…๊ณ„ ๋…ผ์˜๋ฅผ ๋ณด๋ฉด ๋ฐ์ดํ„ฐ ์ค€๋น„๋„๋Š” AI ๊ฐ€์น˜๋ฅผ ์‹คํ˜„ํ•˜๋Š” ๋ฐ ๊ฐ€์žฅ ํฐ ์žฅ์•  ์š”์ธ ์ค‘ ํ•˜๋‚˜๋กœ ๊พธ์ค€ํžˆ ์ง€๋ชฉ๋˜๋ฉฐ, ์ข…์ข… ์ปดํ“จํŒ… ๋น„์šฉ์ด๋‚˜ ๋ชจ๋ธ ์„ ํƒ๋ณด๋‹ค ๋” ํฐ ๋ฌธ์ œ๋กœ ์ธ์‹๋œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ๋˜ํ•œ โ€œ๋Œ€๋ถ€๋ถ„์˜ ๊ธฐ์—…์€ ์—ฌ์ „ํžˆ ๋ถ„์ ˆ๋œ ์‹œ์Šคํ…œ๊ณผ ์ผ๊ด€์„ฑ ์—†๋Š” ๊ฑฐ๋ฒ„๋„Œ์Šค, ๊ทธ๋ฆฌ๊ณ  ์‹ค์ œ๋กœ ์–ด๋–ค ๋ฐ์ดํ„ฐ๋ฅผ ๋ณด์œ ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ ๊ทธ ๋ฐ์ดํ„ฐ๊ฐ€ ์–ผ๋งˆ๋‚˜ ์‹ ๋ขฐํ•  ์ˆ˜ ์žˆ๋Š”์ง€์— ๋Œ€ํ•œ ์ œํ•œ์ ์ธ ๊ฐ€์‹œ์„ฑ ๋ฌธ์ œ๋ฅผ ์•ˆ๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

๋Ÿฌ๋ธ”๋ฆฌ๋Š” IBM ์—ญ์‹œ ๊ณผ๊ฑฐ์—๋Š” ์ด๋Ÿฌํ•œ ๋ฌธ์ œ๋ฅผ ๋‹ค์ˆ˜ ๊ฒช์—ˆ์ง€๋งŒ, ์ง€๋‚œ 3๋…„๊ฐ„ ๋ฐ์ดํ„ฐ๋ฅผ AI์— ์ ํ•ฉํ•œ ์ƒํƒœ๋กœ ๋งŒ๋“ค๊ธฐ ์œ„ํ•ด ์ด๋ฅผ ํ•ด๊ฒฐํ•˜๋Š” ๋ฐ ์ง‘์ค‘ํ•ด ์™”๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

AI ์‹œ๋Œ€๋ฅผ ๊ฒจ๋ƒฅํ•œ IBM์˜ ๋ฐ์ดํ„ฐ ์ „๋žต์€ ์˜ค๋žœ ๊ธฐ๊ฐ„ ์œ ์ง€ํ•ด ์˜จ ๊ธฐ์กด ์ ‘๊ทผ ๋ฐฉ์‹์— ๋Œ€ํ•œ ์—ฌ๋Ÿฌ ๋ณ€ํ™”๋ฅผ ํฌํ•จํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์ด๋ฅผ ํ†ตํ•ด ๋Ÿฌ๋ธ”๋ฆฌ๊ฐ€ ๋งํ•˜๋Š” ํ†ตํ•ฉ ์ „์‚ฌ ๋ฐ์ดํ„ฐ ์•„ํ‚คํ…์ฒ˜๋ฅผ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ์—ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด IBM์€ ๋ฐ์ดํ„ฐ ์˜ค๋„ˆ๋ผ๋Š” ๊ฐœ๋…์„ ์œ ์ง€ํ•˜๋ฉด์„œ๋„, ํ•ด๋‹น ๋ฐ์ดํ„ฐ๊ฐ€ ํŠน์ • ์กฐ์ง์ด ์•„๋‹Œ IBM ์ „์ฒด์˜ ์ž์‚ฐ์ด๋ผ๋Š” ์ ์„ ์ดํ•ดํ•˜๋„๋ก ํ–ˆ๋‹ค. ๋Ÿฌ๋ธ”๋ฆฌ๋Š” โ€œ๋ฐ์ดํ„ฐ๋ฅผ ํ†ต์ œ๋˜๊ณ  ์•ˆ์ „ํ•œ ๋ฐฉ์‹์œผ๋กœ ๋ฏผ์ฃผํ™”ํ•  ์ˆ˜ ์žˆ๋‹ค๋ฉด, ๋น„์ฆˆ๋‹ˆ์Šค๋ฅผ ๋” ํšจ์œจ์ ์ด๊ณ  ์ƒ์‚ฐ์ ์œผ๋กœ ์šด์˜ํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ทธ ๊ฒฐ๊ณผ IBM์€ ์—ฌ๋Ÿฌ ํŒ€์ด ์‚ฌ์ผ๋กœ ํ˜•ํƒœ๋กœ ๋ฐ์ดํ„ฐ๋ฅผ ๊ด€๋ฆฌํ•˜๋˜ ๊ตฌ์กฐ์—์„œ ๋ฒ—์–ด๋‚˜, ๊ณตํ†ต ํ‘œ์ค€๊ณผ ๊ณตํ†ต ์•„ํ‚คํ…์ฒ˜๋ฅผ ์‚ฌ์šฉํ•˜๋Š” ๋‹จ์ผ ์กฐ์ง ์ฒด๊ณ„๋กœ ์ „ํ™˜ํ–ˆ๋‹ค. ๋˜ํ•œ ์ „์‚ฌ ๋ฆฌ๋”๋Š” 300ํ…Œ๋ผ๋ฐ”์ดํŠธ ๊ทœ๋ชจ์˜ ๋ฐ์ดํ„ฐ๋ฅผ ํ†ตํ•ฉํ•˜๊ณ , ํšŒ์‚ฌ๊ฐ€ ์ถ”๊ตฌํ•˜๋Š” ์„ฑ๊ณผ์™€ ์ด๋ฅผ ์ด๋„๋Š” ์›Œํฌํ”Œ๋กœ๋ฅผ ๊ธฐ์ค€์œผ๋กœ ํ•„์š”ํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ์„ ๋ณ„ํ–ˆ๋‹ค.

๋Ÿฌ๋ธ”๋ฆฌ๋Š” โ€œ์˜๋„์ ์œผ๋กœ ์ ‘๊ทผํ–ˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ•˜๋ฉฐ, ํ˜„์žฌ IBM์˜ ๋ฐ์ดํ„ฐ ํ”Œ๋žซํผ์ด ์ „์ฒด ์›Œํฌํ”Œ๋กœ์˜ ์•ฝ 80%๋ฅผ ํฌ๊ด„ํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค. ๊ทธ๋Š” ์ด์–ด โ€œ์˜ค๋Š˜๋‚  ๊ธฐ์—…์—์„œ ๊ฐ€์žฅ ํฐ ์ƒ์‚ฐ์„ฑ ํ–ฅ์ƒ ์š”์ธ ๊ฐ€์šด๋ฐ ํ•˜๋‚˜๋Š” ํ†ตํ•ฉ ์ „์‚ฌ ๋ฐ์ดํ„ฐ ์•„ํ‚คํ…์ฒ˜๋ฅผ ๊ตฌ์ถ•ํ•˜๋Š” ๊ฒƒโ€์ด๋ผ๋ฉฐ โ€œ๋ฐ์ดํ„ฐ์— ๋Œ€ํ•œ ํˆฌ์ž๊ฐ€ ์žˆ์—ˆ๊ธฐ์— IBM์€ AI๋ฅผ ์ „์‚ฌ์ ์œผ๋กœ ๋น ๋ฅด๊ฒŒ ๋„์ž…ํ•  ์ˆ˜ ์žˆ์—ˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

๋” ๋‚˜์€ ๋ฐ์ดํ„ฐ ์ „๋žต์„ ๊ตฌ์ถ•ํ•˜๊ธฐ ์œ„ํ•œ 8๊ฐ€์ง€ ์ œ์–ธ

๋ฐ์ดํ„ฐ ๊ธฐ๋ฐ˜๊ณผ ๋ฐ์ดํ„ฐ ํ™œ์šฉ ์—ญ๋Ÿ‰์—์„œ ๋†’์€ ์„ฑ์ˆ™๋„๋ฅผ ๋‹ฌ์„ฑํ•˜๋ ค๋ฉด, ์กฐ์ง์€ AI ์‹œ๋Œ€์— ๋งž๋Š” ๋ฐ์ดํ„ฐ ์ „๋žต์ด ํ•„์š”ํ•˜๋‹ค. ์ด๋Š” ๋ฐ์ดํ„ฐ ํ’ˆ์งˆ์„ ๊ฐ•ํ™”ํ•˜๊ณ  ๋ฐ์ดํ„ฐ ์‚ฌ์ผ๋กœ๋ฅผ ํ•ด์†Œํ•˜๋ฉฐ, ๋น„์ฆˆ๋‹ˆ์Šค๊ฐ€ ์šฐ์„ ์ˆœ์œ„๋กœ ์‚ผ์€ AI ํ™œ์šฉ ์‚ฌ๋ก€์— ๋ฐ์ดํ„ฐ ์—ญ๋Ÿ‰์„ ์ •๋ ฌํ•˜๋Š” ์ „๋žต์ด๋‹ค.

์—…๊ณ„ ์ „๋ฌธ๊ฐ€๋Š” ๋‹ค์Œ๊ณผ ๊ฐ™์€ ์‹คํ–‰ ๋‹จ๊ณ„๋ฅผ ์ œ์‹œํ•œ๋‹ค.

1. ๋ฐ์ดํ„ฐ ์†Œ์œ  ๊ฐœ๋…์„ ์žฌ์ •์˜ํ•˜๋ผ

IDC์˜ ๋ผ์ดํŠธ๋Š” โ€œ๋ฐ์ดํ„ฐ ์†Œ์œ ๋ฅผ ์ˆœ์ˆ˜ํ•œ IT ๋ฌธ์ œ๋กœ๋งŒ ๋‹ค๋ฃจ๋Š” ์ „ํ†ต์  ๋ชจ๋ธ์€ ๋” ์ด์ƒ ์œ ํšจํ•˜์ง€ ์•Š๋‹คโ€๋ผ๋ฉฐ โ€œ์‚ฌ์—…๋ถ€, ์ œํ’ˆ ์กฐ์ง, AI ํ”Œ๋žซํผ์ด ์ง€์†์ ์œผ๋กœ ๋ฐ์ดํ„ฐ๋ฅผ ์ƒ์„ฑํ•˜๊ณ  ๋ณ€ํ™˜ํ•˜๋Š” ํ™˜๊ฒฝ์—์„œ๋Š” ์ด๋Ÿฌํ•œ ์ ‘๊ทผ์ด ์ž‘๋™ํ•˜์ง€ ์•Š๋Š”๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ์ด์ƒ์ ์œผ๋กœ๋Š” CDO์™€ ๊ฐ™์€ ๊ณ ์œ„ ๋ฐ์ดํ„ฐ ๋ฆฌ๋”๊ฐ€ ๋ช…ํ™•ํ•œ ์ฑ…์ž„์„ ์ ธ์•ผ ํ•˜์ง€๋งŒ, CDO๊ฐ€ ์—†๋Š” ์กฐ์ง์ด๋ผ๋ฉด IT, ๋ณด์•ˆ, ์‚ฌ์—… ๋ถ€๋ฌธ ์ „๋ฐ˜์— ๊ฑธ์ณ ๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค ์ฑ…์ž„์„ ๋ช…ํ™•ํ•˜๊ฒŒ ๋ถ„์‚ฐํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

๋˜ํ•œ ๊ทธ๋Š” โ€œ์ •์ฑ…์„ ์ •์˜ํ•˜๋Š” ๋‹จ์ผํ•œ ๊ถŒํ•œ ์ฐฝ๊ตฌ์™€ ์‹คํ–‰์„ ๋‹ด๋‹นํ•˜๋Š” ์—ฐํ•ฉํ˜• ๋ชจ๋ธ์„ ํ•จ๊ป˜ ๊ฐ–์ถ”๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•˜๋‹คโ€๋ผ๋ฉฐ โ€œ์ด๋ฅผ ํ†ตํ•ด ์‚ฌ์—…๋ถ€์˜ ์ž์œจ์„ฑ์€ ์œ ์ง€ํ•˜๋˜, ํ†ต์ œ๋˜์ง€ ์•Š์€ ์ƒํƒœ๋Š” ํ”ผํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ๋ง๋ถ™์˜€๋‹ค.

์„ธ์ธํŠธํ† ๋จธ์Šค๋Œ€ํ•™๊ต ์†Œํ”„ํŠธ์›จ์–ด๊ณตํ•™ยท๋ฐ์ดํ„ฐ์‚ฌ์ด์–ธ์Šคํ•™๊ณผ ํ•™๊ณผ์žฅ์ด์ž ์‘์šฉ ์ธ๊ณต์ง€๋Šฅ ์„ผํ„ฐ ์†Œ์žฅ์ธ ๋งŒ์ง“ ๋ ˆ๊ฒŒ๋Š” ๋ฐ์ดํ„ฐ ์˜ค๋„ˆ์˜ ์—ญํ• ์„ ๋ฐ์ดํ„ฐ ๊ด€๋ฆฌ์ž, ์ฆ‰ ๋ฐ์ดํ„ฐ ์ŠคํŠœ์–ด๋“œ๋กœ ์žฌ์ •์˜ํ•  ๊ฒƒ์„ ๊ถŒ๊ณ ํ–ˆ๋‹ค. ๋ ˆ๊ฒŒ๋Š” ์ด๋“ค์ด ๋ฐ์ดํ„ฐ๋ฅผ ์†Œ์œ ํ•˜๋Š” ์กด์žฌ๊ฐ€ ์•„๋‹ˆ๋ผ, ์ค‘์•™ ๋ฐ์ดํ„ฐ ์กฐ์ง์ด ์ •ํ•œ ํ‘œ์ค€๊ณผ ๊ฑฐ๋ฒ„๋„Œ์Šค, ๋ณด์•ˆ, ์ƒํ˜ธ์šด์šฉ์„ฑ์„ ๊ธฐ๋ฐ˜์œผ๋กœ ๋ฐ์ดํ„ฐ์˜ ์˜๋ฏธ์™€ ํ’ˆ์งˆ์„ ์ฑ…์ž„์ง€๋Š” ์—ญํ• ์„ ๋งก์•„์•ผ ํ•œ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

2. ๋ฐ์ดํ„ฐ ์‚ฌ์ผ๋กœ๋ฅผ ํ•ด์†Œํ•˜๋ผ

์ด๋ฅผ ์œ„ํ•ด ๋ผ์ดํŠธ๋Š” โ€œ์ƒ์„ฑํ˜• AI๋Š” ์›Œํฌํ”Œ๋กœ์™€ ํ”„๋กœ์„ธ์Šค, ๋ฐ์ดํ„ฐ ์†Œ์Šค๊ฐ€ ์ „์‚ฌ์ ์œผ๋กœ ์—ฐ๊ฒฐ๋  ๋•Œ์—๋งŒ ๊ฐ€์น˜๋ฅผ ์ฐฝ์ถœํ•˜๊ธฐ ๋•Œ๋ฌธ์—, CIO๋Š” ๊ณตํ†ต์˜ AI ๋ฐ ๋ฐ์ดํ„ฐ ์„ฑ๊ณผ๋ฅผ ์ค‘์‹ฌ์œผ๋กœ ์‚ฌ์—… ๋ถ€๋ฌธ์„ ์ •๋ ฌํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ทธ๋Š” ์ด์–ด โ€œ์ด๋ฅผ ์œ„ํ•ด์„œ๋Š” ํฌ๋กœ์ŠคํŽ‘์…”๋„ ๊ฑฐ๋ฒ„๋„Œ์Šค๋ฅผ ๊ตฌ์ถ•ํ•˜๊ณ , ๋ถ„๋ฅ˜ ์ฒด๊ณ„์™€ ์ •์ฑ…์„ ํ‘œ์ค€ํ™”ํ•˜๋ฉฐ, ๋ฐ์ดํ„ฐ๋ฅผ ๋ณดํ˜ธํ•˜๊ธฐ๋ณด๋‹ค ๊ณต์œ ํ•˜๋„๋ก ์œ ๋„ํ•˜๋Š” ์ธ์„ผํ‹ฐ๋ธŒ๋ฅผ ๋งˆ๋ จํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ๋˜ํ•œ โ€œํ†ตํ•ฉ ํ”Œ๋žซํผ, ๋ฉ”ํƒ€๋ฐ์ดํ„ฐ ๊ณ„์ธต, ๊ณตํ†ต ๋ณด์•ˆ ํ”„๋ ˆ์ž„์›Œํฌ ๊ฐ™์€ ๊ธฐ์ˆ ๋„ ๋„์›€์ด ๋˜์ง€๋งŒ, ์ง„์ •ํ•œ ์ „ํ™˜์„ ์ด๋„๋Š” ์š”์†Œ๋Š” C๋ ˆ๋ฒจ๊ณผ ์ฃผ์š” ๋น„์ฆˆ๋‹ˆ์Šค ์ดํ•ด๊ด€๊ณ„์ž ์ „๋ฐ˜์— ๊ฑธ์นœ ์กฐ์œจ๋œ ๋ฆฌ๋”์‹ญโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

3. AI ์‹œ๋Œ€๋ฅผ ์œ„ํ•œ ๋ฐ์ดํ„ฐ ๊ธฐ์ˆ ์— ํˆฌ์žํ•˜๋ผ

๋ผ์ดํŠธ๋Š” AI ์‹œ๋Œ€์— ํ•„์š”ํ•œ ๋ฐ์ดํ„ฐ ๊ธฐ์ˆ ๋กœ ํ˜„๋Œ€์ ์ธ ๋ฐ์ดํ„ฐ ๋ ˆ์ดํฌ์™€ ๋ฐ์ดํ„ฐ ๋ ˆ์ดํฌํ•˜์šฐ์Šค, ๋ฒกํ„ฐ ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค, ํ™•์žฅํ˜• ์˜ค๋ธŒ์ ํŠธ ์Šคํ† ๋ฆฌ์ง€๋ฅผ ๊ผฝ์•˜๋‹ค. ๊ทธ๋Š” ์ด๋Ÿฌํ•œ ๊ธฐ์ˆ ์ด โ€œ๊ฐ•๋ ฅํ•œ ๊ฑฐ๋ฒ„๋„Œ์Šค๋ฅผ ์œ ์ง€ํ•˜๋ฉด์„œ๋„ ๋Œ€์šฉ๋Ÿ‰์˜ ๋ฉ€ํ‹ฐ๋ชจ๋‹ฌ ๋ฐ์ดํ„ฐ๋ฅผ ์ฒ˜๋ฆฌํ•  ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋˜ํ•œ ์กฐ์ง์€ AI ์›Œํฌํ”Œ๋กœ๊ฐ€ ์ฒ˜์Œ๋ถ€ํ„ฐ ๋๊นŒ์ง€ ์•ˆ์ •์ ์œผ๋กœ ์ž‘๋™ํ•˜๋„๋ก ๋ฐ์ดํ„ฐ ์ˆ˜์ง‘, ์ •์ œ, ๋ณ€ํ™˜, ์ด๋™์„ ์ž๋™ํ™”ํ•˜๋Š” ์˜ค์ผ€์ŠคํŠธ๋ ˆ์ด์…˜ ๋ฐ ํŒŒ์ดํ”„๋ผ์ธ ๋„๊ตฌ๋ฅผ ๊ฐ–์ถฐ์•ผ ํ•œ๋‹ค. ๋ชจ๋ธ์ด ๋ฐ์ดํ„ฐ์˜ ๋งฅ๋ฝ์„ ์ดํ•ดํ•˜๊ณ  ๊ณ„๋ณด๋ฅผ ์ถ”์ ํ•˜๋ฉฐ ๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ์™€ ๋น„๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ๋ฅผ ์•ˆ์ „ํ•˜๊ณ  ์‹ ๋ขฐ์„ฑ ์žˆ๊ฒŒ ํ™œ์šฉํ•˜๋ ค๋ฉด ๋ฉ”ํƒ€๋ฐ์ดํ„ฐ ์—”์ง„๊ณผ ๊ฑฐ๋ฒ„๋„Œ์Šค ๊ณ„์ธต ์—ญ์‹œ ํ•„์ˆ˜์ ์ด๋‹ค.

๋งŒ์ง“ ๋ ˆ๊ฒŒ๋Š” โ€œ๋ชจ๋“ˆํ˜•์ด๋ฉฐ ๊ฑฐ๋ฒ„๋„Œ์Šค๊ฐ€ ์ ์šฉ๋˜๊ณ , ์ง€์†์ ์œผ๋กœ ์ง„ํ™”ํ•  ์ˆ˜ ์žˆ๋Š” ๋ฐ์ดํ„ฐ ํ”Œ๋žซํผ ๊ณ„์ธต์„ ๊ตฌ์ถ•ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์กฐ์–ธํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ๋ฐ์ดํ„ฐ๋ฅผ ๋‹จ์ผ ํŒŒ์ดํ”„๋ผ์ธ์„ ์œ„ํ•œ ์ž์›์ด ์•„๋‹ˆ๋ผ ์žฌ์‚ฌ์šฉ ๊ฐ€๋Šฅํ•œ ์ œํ’ˆ์œผ๋กœ ๋‹ค๋ฃฐ ์ˆ˜ ์žˆ์–ด์•ผ ํ•˜๋ฉฐ, ๋ฐฐ์น˜ ์ฒ˜๋ฆฌ์™€ ์‹ค์‹œ๊ฐ„ ์š”๊ตฌ๋ฅผ ๋ชจ๋‘ ์ง€์›ํ•  ์ˆ˜ ์žˆ๋Š” ์•„ํ‚คํ…์ฒ˜๊ฐ€ ํ•„์š”ํ•˜๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ ˆ๊ฒŒ๋Š” ๋ฐ์ดํ„ฐ ๋ ˆ์ดํฌ์™€ ๋ฐ์ดํ„ฐ ๋ ˆ์ดํฌํ•˜์šฐ์Šค์— ๋Œ€ํ•ด์„œ๋„ ๊ธ์ •์ ์ธ ํ‰๊ฐ€๋ฅผ ๋‚ด๋ ธ๋‹ค. ๊ทธ๋Š” ์ด๋“ค ๊ธฐ์ˆ ์ด ๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ์™€ ๋น„๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ๋ฅผ ๋ชจ๋‘ ์ฒ˜๋ฆฌํ•  ์ˆ˜ ์žˆ์–ด โ€œAI์˜ ํ•ต์‹ฌ ๊ธฐ๋ฐ˜์œผ๋กœ ์ž๋ฆฌ ์žก๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

๋˜ํ•œ ์˜ํŠธ์›์Šค์˜ ์ตœ๊ณ  AIยท๋ฐ์ดํ„ฐ ์ฑ…์ž„์ž์ธ ์ƒค์–€ ๋ชจํ•œํ‹ฐ๋Š” CIO์—๊ฒŒ ๋ชจ๋“ˆํ˜• ๊ธฐ์ˆ ๊ณผ ์œ ์—ฐํ•œ ๊ตฌ์กฐ๋ฅผ ๊ฐ–์ถ˜ ์ปดํฌ์ €๋ธ” ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ๋ฅผ ๊ตฌ์ถ•ํ•  ๊ฒƒ์„ ๊ถŒ๊ณ ํ–ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ์‚ฌ๋žŒ๊ณผ AI๊ฐ€ ์—ฌ๋Ÿฌ ๊ณ„์ธต์— ๊ฑธ์ณ ๋ฐ์ดํ„ฐ์— ์ ‘๊ทผํ•˜๊ณ  ํ˜‘์—…ํ•  ์ˆ˜ ์žˆ๋‹ค๋Š” ์„ค๋ช…์ด๋‹ค.

์ „๋ฌธ๊ฐ€๋“ค์€ ๋ฐ์ดํ„ฐ ๋ผ์ดํ”„์‚ฌ์ดํด์˜ ๋ณ€ํ™”์— ๋Œ€์‘ํ•˜๋Š” ๊ธฐ์ˆ  ํˆฌ์ž๋„ ์ค‘์š”ํ•˜๋‹ค๊ณ  ์กฐ์–ธํ•œ๋‹ค.

๋ผ์ดํŠธ๋Š” โ€œ์ƒ์„ฑํ˜• AI๋Š” ๋ฐ์ดํ„ฐ ๋ผ์ดํ”„์‚ฌ์ดํด์„ ๊ทผ๋ณธ์ ์œผ๋กœ ์žฌํŽธํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์ผ์‹œ์  ๋ฐ์ดํ„ฐ, ์บ์‹œ ๋ฐ์ดํ„ฐ, ์˜๊ตฌ ์ €์žฅ ๋ฐ์ดํ„ฐ๊ฐ€ ํ›จ์”ฌ ๋” ์—ญ๋™์ ์œผ๋กœ ํ˜ผ์žฌ๋˜๋Š” ํ™˜๊ฒฝ์„ ๋งŒ๋“ค๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ๋Œ€๋ถ€๋ถ„์˜ ์ƒ์„ฑํ˜• AI ๊ฒฐ๊ณผ๋ฌผ์€ ์ˆ˜ ์ดˆ์—์„œ ์ˆ˜ ๋ถ„, ๊ธธ์–ด์•ผ ์ˆ˜ ์‹œ๊ฐ„๋งŒ ์‚ฌ์šฉ๋˜๊ธฐ ๋•Œ๋ฌธ์—, ๋น ๋ฅธ ๋ฐ˜๋ณต๊ณผ ์บ์‹ฑ, ๋ณ€๋™์„ฑ์ด ํฐ ์›Œํฌํ”Œ๋กœ๋ฅผ ์ฒ˜๋ฆฌํ•  ์ˆ˜ ์žˆ๋Š” DRAM๊ณผ SSD ๊ฐ™์€ ๊ณ ์„ฑ๋Šฅ ์ธํ”„๋ผ์— ๋Œ€ํ•œ ์ˆ˜์š”๊ฐ€ ์ปค์ง€๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

๋ฐ˜๋ฉด ๊ทธ๋Š” โ€œ์ตœ์ข… ๋ฌธ์„œ, ์Šน์ธ๋œ ๋ฏธ๋””์–ด ์ž์‚ฐ, ํ•ฉ์„ฑ ํ•™์Šต ๋ฐ์ดํ„ฐ, ๊ทœ์ œ ๋Œ€์‘๊ณผ ๊ด€๋ จ๋œ ์ฝ˜ํ…์ธ  ๋“ฑ ์ผ๋ถ€ ์ƒ์„ฑํ˜• AI ๊ฒฐ๊ณผ๋ฌผ์€ ์žฅ๊ธฐ๊ฐ„ ๋ณด์กด๋œ๋‹คโ€๋ผ๋ฉฐ โ€œ์ด๋Ÿฌํ•œ ๋ฐ์ดํ„ฐ๋Š” ์—ฌ์ „ํžˆ ๋น„์šฉ ํšจ์œจ์ ์ด๋ฉด์„œ ๋Œ€์šฉ๋Ÿ‰์„ ์ œ๊ณตํ•˜๋Š” HDD ๊ธฐ๋ฐ˜ ์Šคํ† ๋ฆฌ์ง€์— ํฌ๊ฒŒ ์˜์กดํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ด์–ด โ€œ์ƒ์„ฑํ˜• AI ๋„์ž…์ด ํ™•๋Œ€๋ ์ˆ˜๋ก, ์ดˆ๊ณ ์† ๋ฉ”๋ชจ๋ฆฌ๋ฅผ ํ™œ์šฉํ•œ ์ผ์‹œ์  ์ฝ˜ํ…์ธ  ์ฒ˜๋ฆฌ๋ถ€ํ„ฐ HDD ๊ธฐ๋ฐ˜์˜ ๊ฒฌ๊ณ ํ•œ ์•„์นด์ด๋ธŒ๊นŒ์ง€ ์ „ ์ฃผ๊ธฐ๋ฅผ ํฌ๊ด„ํ•˜๋Š” ๋ฐ์ดํ„ฐ ์ „๋žต์ด ํ•„์š”ํ•ด์งˆ ๊ฒƒโ€์ด๋ผ๋ฉฐ โ€œ์Šคํ† ๋ฆฌ์ง€ ๋ถ€๋‹ด๊ณผ ๊ตฌ์กฐ ์ž์ฒด๊ฐ€ ๋ณ€ํ™”ํ•˜๊ณ  ์žˆ๊ธฐ ๋•Œ๋ฌธโ€์ด๋ผ๊ณ  ๋ถ„์„ํ–ˆ๋‹ค.

4. ๋ฐ์ดํ„ฐ ์•„ํ‚คํ…์ฒ˜์— ์ž๋™ํ™”์™€ ์ง€๋Šฅ ์š”์†Œ๋ฅผ ๋”ํ•˜๋ผ

๋ชจํ•œํ‹ฐ๋Š” ๊ธฐ์—… ๋ฐ์ดํ„ฐ ํ™˜๊ฒฝ์ด ์—ด์•…ํ•œ ์›์ธ์œผ๋กœ ๋ฐ์ดํ„ฐ ์ƒ์‚ฐ์ž์™€ ๋ฐ์ดํ„ฐ ์†Œ๋น„์ž ๊ฐ„์˜ ๋‹จ์ ˆ์„ ์ง€๋ชฉํ–ˆ๋‹ค. ๊ทธ๋Š” ์ƒ์„ฑ๋œ ๋ฐ์ดํ„ฐ๊ฐ€ ์ด๋ฅธ๋ฐ” ๋ฐ์ดํ„ฐ ์›จ์–ดํ•˜์šฐ์Šค๋ผ๋Š” โ€œ์–ด๋”˜๊ฐ€์˜ ๊ฑฐ๋Œ€ํ•œ ๋”๋ฏธโ€์— ์Œ“์ธ ๋’ค, ์ด๋ฅผ ํ™œ์šฉํ•˜๊ธฐ ์œ„ํ•ด ๋ณ„๋„์˜ ๋ถ„์„ ๊ณ„์ธต์„ ๋ง๋ถ™์ด๋Š” ๋ฐฉ์‹์ด ์ผ๋ฐ˜์ ์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ด๋Ÿฌํ•œ ์ ‘๊ทผ์€ ์‹ค์ œ๋กœ ์ž‘๋™ํ•˜๊ฒŒ ๋งŒ๋“ค๊ธฐ ์œ„ํ•ด ๋งŽ์€ ์ธ์  ์ง€์‹๊ณผ ์ˆ˜์ž‘์—…์„ ํ•„์š”๋กœ ํ•œ๋‹ค๋Š” ์ง€์ ์ด๋‹ค.

์ด์— ๋”ฐ๋ผ ๋ชจํ•œํ‹ฐ๋Š” ๋ฐ์ดํ„ฐ ์ƒ์‚ฐ์ž์™€ ๋ฐ์ดํ„ฐ ์†Œ๋น„์ž์˜ ๊ฑฐ๋ฆฌ๋ฅผ ์ขํžˆ๊ธฐ ์œ„ํ•ด ๋ฐ์ดํ„ฐ ์ œํ’ˆ ๊ด€์ ์„ ๋„์ž…ํ•˜๊ณ , ํ•„์š”ํ•  ๋•Œ AI๊ฐ€ ์ ์ ˆํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ์‹๋ณ„ํ•˜๊ณ  ์ ‘๊ทผํ•  ์ˆ˜ ์žˆ๋„๋ก ์ „์‚ฌ ์•„ํ‚คํ…์ฒ˜์— ์ž๋™ํ™”์™€ ์ง€๋Šฅ์„ ์ถ”๊ฐ€ํ•ด์•ผ ํ•œ๋‹ค๊ณ  ์กฐ์–ธํ–ˆ๋‹ค.

๊ทธ๋Š” CIO๊ฐ€ MCP(Model Context Protocol)์„ ํ™œ์šฉํ•ด ๋ฐ์ดํ„ฐ๋ฅผ ๊ฐ์‹ธ๊ณ , ํ”„๋กœํ† ์ฝœ ์ˆ˜์ค€์˜ ์ ‘๊ทผ ๋ฐฉ์‹์„ ์ œ๊ณตํ•  ์ˆ˜ ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๋‹ค๋งŒ ์ด๋ฅผ ์œ„ํ•ด์„œ๋Š” ๋ฐ์ดํ„ฐ์˜ ๋ฐœ๊ฒฌ ๊ฐ€๋Šฅ์„ฑ์„ ๋ณด์žฅํ•  ์ˆ˜ ์žˆ๋„๋ก, ์กฐ์ง์ด ๋ฐ์ดํ„ฐ ์นดํƒˆ๋กœ๊ทธ์™€ ๋„๊ตฌ ์ „๋ฐ˜์— ๊ฑธ์ณ ๊ด€๋ จ ์ •๋ณด๋ฅผ ์ฒด๊ณ„์ ์œผ๋กœ ์ธ์ฝ”๋”ฉํ•ด์•ผ ํ•œ๋‹ค๊ณ  ๋ง๋ถ™์˜€๋‹ค.

5. ๊ตฌ์กฐํ™”ยท๋น„๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ๋ฅผ AI์— ๋งž๊ฒŒ ์ค€๋น„ํ•˜๋ผ

๋ผ์ดํŠธ๋Š” โ€œ๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ๋Š” ์ผ๊ด€๋œ ํ˜•์‹์œผ๋กœ ์ •๋ฆฌ๋˜๊ณ , ์ ์ ˆํ•œ ๊ฑฐ๋ฒ„๋„Œ์Šค๊ฐ€ ์ ์šฉ๋˜๋ฉฐ, ์ •ํ™•ํ•œ ๋ฉ”ํƒ€๋ฐ์ดํ„ฐ๋กœ ๋ณด๊ฐ•๋  ๋•Œ AI์— ์ ํ•ฉํ•œ ์ƒํƒœ๊ฐ€ ๋œ๋‹คโ€๋ผ๋ฉฐ โ€œ์ด ๊ฒฝ์šฐ ๋ชจ๋ธ์ด ๋ฐ์ดํ„ฐ๋ฅผ ์ดํ•ดํ•˜๊ณ  ํ™œ์šฉํ•˜๊ธฐ๊ฐ€ ํ›จ์”ฌ ์‰ฌ์›Œ์ง„๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ์กฐ์ง์€ ๊ฐ•๋ ฅํ•œ ๋ฐ์ดํ„ฐ ํ’ˆ์งˆ ๊ด€๋ฆฌ์™€ ๋งˆ์Šคํ„ฐ ๋ฐ์ดํ„ฐ ๊ด€๋ฆฌ, ๋ช…ํ™•ํ•œ ์ฑ…์ž„ ์ฒด๊ณ„๋ฅผ ์šฐ์„ ์ ์œผ๋กœ ๊ตฌ์ถ•ํ•ด ๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ์…‹์ด ์‹ ๋ขฐ์„ฑ๊ณผ ์ƒํ˜ธ์šด์šฉ์„ฑ์„ ์œ ์ง€ํ•˜๊ณ , ํŠน์ • AI ํ™œ์šฉ ์‚ฌ๋ก€์— ๋งž๊ฒŒ ์ •๋ ฌ๋˜๋„๋ก ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

์ „๋ฌธ๊ฐ€๋“ค์€ ์ด๋Ÿฌํ•œ ๊ทœ์œจ์„ ๋น„๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ์—๋„ ๋™์ผํ•˜๊ฒŒ ์ ์šฉํ•ด์•ผ ํ•œ๋‹ค๊ณ  ๊ฐ•์กฐํ•œ๋‹ค. ๋น„๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ ์—ญ์‹œ ์ ์ ˆํ•œ ํƒœ๊น…๊ณผ ๋ถ„๋ฅ˜, ๋ฉ”ํƒ€๋ฐ์ดํ„ฐ ๋ณด๊ฐ•์„ ํ†ตํ•ด AI ์‹œ์Šคํ…œ์ด ํšจ๊ณผ์ ์œผ๋กœ ์ดํ•ดํ•˜๊ณ  ๊ฒ€์ƒ‰ํ•  ์ˆ˜ ์žˆ๋„๋ก ์ค€๋น„๋ผ์•ผ ํ•œ๋‹ค๋Š” ์„ค๋ช…์ด๋‹ค.

๋งŒ์ง“ ๋ ˆ๊ฒŒ๋Š” โ€œ๋น„๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ๋ฅผ ์ผ๊ธ‰ ๋ฐ์ดํ„ฐ ์ž์‚ฐ์œผ๋กœ ๋‹ค๋ค„์•ผ ํ•œ๋‹คโ€๋ผ๋ฉฐ โ€œ๊ณ ๊ฐ ์„œ๋น„์Šค ์Œ์„ฑ ํ†ตํ™”, ๋ฉ”์‹œ์ง€, ๋ฌธ์„œ์™€ ๊ฐ™์€ ๋น„๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ์— ๊ฐ€์žฅ ํฅ๋ฏธ๋กœ์šด AI ํ™œ์šฉ ์‚ฌ๋ก€๊ฐ€ ์กด์žฌํ•˜์ง€๋งŒ, ๋งŽ์€ ์กฐ์ง์—์„œ ์—ฌ์ „ํžˆ ๋น„๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ๋Š” ์‚ฌ๊ฐ์ง€๋Œ€๋กœ ๋‚จ์•„ ์žˆ๋‹คโ€๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค.

๋ ˆ๊ฒŒ๋Š” ์ด๋Ÿฌํ•œ ๋น„๊ตฌ์กฐํ™” ๋ฐ์ดํ„ฐ๋ฅผ ๊ฒ€์ƒ‰ ๊ฐ€๋Šฅํ•œ ํ˜•ํƒœ๋กœ ํ™œ์šฉํ•˜๊ธฐ ์œ„ํ•ด ๋ฒกํ„ฐ ๋ฐ์ดํ„ฐ๋ฒ ์ด์Šค์— ์ €์žฅํ•  ๊ฒƒ์„ ๊ถŒ๊ณ ํ–ˆ๋‹ค.

6. ์™ธ๋ถ€ ๋ฐ์ดํ„ฐ ์†Œ์Šค์™€ ํ•ฉ์„ฑ ๋ฐ์ดํ„ฐ ํ™œ์šฉ์„ ๊ฒ€ํ† ํ•˜๋ผ

๋ผ์ดํŠธ๋Š” โ€œ๊ธฐ์กด ๋ฐ์ดํ„ฐ๊ฐ€ ๋ถˆ์™„์ „ํ•˜๊ฑฐ๋‚˜ ํŽธํ–ฅ๋ผ ์žˆ๊ฑฐ๋‚˜, ๊ทœ๋ชจ๊ฐ€ ๋ถ€์กฑํ•˜๊ฑฐ๋‚˜, ์ถ”์ง„ํ•˜๋ ค๋Š” AI ํ™œ์šฉ ์‚ฌ๋ก€์™€ ์ถฉ๋ถ„ํžˆ ๋งž์ง€ ์•Š๋Š” ๊ฒฝ์šฐ์—๋Š” ์™ธ๋ถ€ ๋ฐ์ดํ„ฐ๋‚˜ ํ•ฉ์„ฑ ๋ฐ์ดํ„ฐ๊ฐ€ ํ•„์š”ํ•œ์ง€ ๋ฐ˜๋“œ์‹œ ๊ฒ€ํ† ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œํ•ฉ์„ฑ ๋ฐ์ดํ„ฐ๋Š” ์‹ค์ œ ๋ฐ์ดํ„ฐ๊ฐ€ ๋ฏผ๊ฐํ•˜๊ฑฐ๋‚˜ ์ˆ˜์ง‘ ๋น„์šฉ์ด ๋†’๊ณ , ๊ฐœ์ธ์ •๋ณด ๋ณดํ˜ธ๋‚˜ ๊ทœ์ œ, ์šด์˜์ƒ์˜ ์ œ์•ฝ์œผ๋กœ ํ™œ์šฉ์ด ์ œํ•œ๋  ๋•Œ ํŠนํžˆ ์œ ์šฉํ•˜๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

7. ๊ณ ์„ฑ์ˆ™๋„ ๋ฐ์ดํ„ฐ ๊ธฐ๋ฐ˜์„ ๋‹จ๊ณ„์ ์œผ๋กœ ๊ตฌ์ถ•ํ•˜๋ผ

์„ธ์ผ์ฆˆํฌ์Šค์˜ ์ „์‚ฌ IT ์ „๋žต ๋‹ด๋‹น ์ˆ˜์„๋ถ€์‚ฌ์žฅ์ธ ์‹œ๋ฐ”๋‹ˆ ์•„ํ›„์ž๋Š” ๋ฐ์ดํ„ฐ๊ฐ€ ์™„๋ฒฝํ•˜๊ฒŒ ์ •๋น„๋  ๋•Œ๊นŒ์ง€ ๊ธฐ๋‹ค๋ฆฌ์ง€ ๋ง๋ผ๊ณ  ์กฐ์–ธํ–ˆ๋‹ค.

์•„ํ›„์ž๋Š” โ€œ๋ชจ๋“  ๋ฐ์ดํ„ฐ๋ฅผ ์™„๋ฒฝํ•˜๊ฒŒ ์ค€๋น„ํ•œ ๋’ค์—์•ผ ๋ณธ๊ฒฉ์ ์œผ๋กœ ์‹œ์ž‘ํ•  ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋А๋ผ๋Š” ์กฐ์ง์ด ์žˆ์ง€๋งŒ, ๋™์‹œ์— AI ์—ฌ์ •์„ ์‹œ์ž‘ํ•˜๋ผ๋Š” ์••๋ฐ•๋„ ๋ฐ›๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ทธ๋Š” ๋Œ€๋ถ€๋ถ„์˜ ์ „์‚ฌ ํ”„๋กœ๊ทธ๋žจ์ด ์„ฑ์ˆ™ํ•ด ๊ฐ€๋Š” ๊ณผ์ •๊ณผ ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ, AI ์‹œ๋Œ€๋ฅผ ์œ„ํ•œ ๋ฐ์ดํ„ฐ ํ”„๋กœ๊ทธ๋žจ ์—ญ์‹œ ์ ์ง„์ ์ธ ์ ‘๊ทผ์ด ํ•„์š”ํ•˜๋‹ค๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค. CIO์™€ ๊ฒฝ์˜์ง„์€ ๋‹จ๊ณ„์ ์œผ๋กœ ๋ฐ์ดํ„ฐ ํ”„๋กœ๊ทธ๋žจ์„ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ๊ณ , ๋˜ ๊ทธ๋ ‡๊ฒŒ ํ•ด์•ผ ํ•œ๋‹ค๋Š” ๊ฒƒ์ด๋‹ค.

์•„ํ›„์ž๋Š” ํ•˜๋‚˜์˜ AI ๊ธฐ๋ฐ˜ ์„ฑ๊ณผ๋ฅผ ์ง€์›ํ•˜๋Š” ๋ฐ์ดํ„ฐ ์ „๋žต๊ณผ ์•„ํ‚คํ…์ฒ˜๋ฅผ ๋จผ์ € ๊ตฌ์ถ•ํ•œ ๋’ค, ์ด๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ๋‹ค์Œ ์„ฑ๊ณผ๋กœ ํ™•์žฅํ•ด ๋‚˜๊ฐ€๋Š” ๋ฐฉ์‹์œผ๋กœ ๋ฐ์ดํ„ฐ ํ”„๋กœ๊ทธ๋žจ์„ ๊ณ ๋„ํ™”ํ•  ๊ฒƒ์„ ๊ถŒ๊ณ ํ–ˆ๋‹ค.

๊ทธ๋Š” โ€œํ•„์š”ํ•œ ๊ฒฐ๊ณผ์—์„œ ๊ฑฐ๊พธ๋กœ ์„ค๊ณ„ํ•˜๋Š” ์‚ฌ๊ณ ๋ฐฉ์‹โ€์ด๋ผ๋ฉฐ โ€œ์šด์˜ ํ™˜๊ฒฝ์— ๋ฐฐํฌํ•˜๊ณ , ์ ์ ˆํ•œ ๊ฐ€๋“œ๋ ˆ์ผ์„ ๊ฐ–์ท„๋Š”์ง€ ํ™•์ธํ•œ ๋’ค ์ด๋ฅผ ๊ด€์ฐฐํ•˜๊ณ  ์กฐ์ •ํ•ด ํ™•์žฅ์„ฑ์„ ํ™•๋ณดํ•œ ๋‹ค์Œ, ๋‹ค์Œ ๋‹จ๊ณ„๋ฅผ ์ง„ํ–‰ํ•˜๋ฉด ๋œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

8. ๋ฐ์ดํ„ฐ ์กฐ์ง์„ ์ „์‚ฌ ํ˜‘์—… ๊ตฌ์กฐ๋กœ ๊ตฌ์ถ•ํ•˜๋ผ

๋ผ์ดํŠธ๋Š” โ€œ๋ฐ์ดํ„ฐ๋Š” IT, ๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค, ๋ณด์•ˆ, ๊ทธ๋ฆฌ๊ณ  ์‹ค์ œ๋กœ ๋ฐ์ดํ„ฐ๋ฅผ ํ™œ์šฉํ•ด ์˜์‚ฌ๊ฒฐ์ •์„ ๋‚ด๋ฆฌ๋Š” ์‚ฌ์—… ๋ถ€๋ฌธ์„ ๋ชจ๋‘ ์•„์šฐ๋ฅด๋Š” ํฌ๋กœ์ŠคํŽ‘์…”๋„ ์ƒํƒœ๊ณ„์˜ ์ง€์›์„ ๋ฐ›์•„์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ทธ๋Š” โ€œAI ์‹œ๋Œ€์˜ ๋ฐ์ดํ„ฐ ์ „๋žต์€ ์ด๋“ค ์กฐ์ง์ด ๊ณต๋™์œผ๋กœ ์ฑ…์ž„์„ ๋‚˜๋ˆŒ ๋•Œ ๊ฐ€์žฅ ํšจ๊ณผ์ ์œผ๋กœ ์ž‘๋™ํ•œ๋‹คโ€๋ผ๋ฉฐ โ€œIT ์กฐ์ง์€ ์ธํ”„๋ผ๋ฅผ ๋’ท๋ฐ›์นจํ•˜๊ณ , ๊ฑฐ๋ฒ„๋„Œ์Šค ์กฐ์ง์€ ์‹ ๋ขฐ์„ฑ๊ณผ ํ’ˆ์งˆ์„ ๋ณด์žฅํ•˜๋ฉฐ, ์‚ฌ์—… ์กฐ์ง์€ ๋ฐ์ดํ„ฐ์˜ ๋งฅ๋ฝ๊ณผ ๊ฐ€์น˜๋ฅผ ์ •์˜ํ•˜๋Š” ์—ญํ• ์„ ๋งก์•„์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋งํ–ˆ๋‹ค.
dl-ciokorea@foundryco.com

Before yesterdayMain stream

DORA Compliance Checklist for Cybersecurity

By: FireMon
15 December 2025 at 16:38

The Digital Operational Resilience Act (DORA) is now in full effect, and financial institutions across the EU face mounting pressure to demonstrate robust ICT risk management and cyber resilience. With...

The post DORA Compliance Checklist for Cybersecurity appeared first on Security Boulevard.

Appeals court judges scrutinize Trumpโ€™s national security basis for collective bargaining rollback

Federal appeals court judges are weighing what limits, if any, exist for President Donald Trump to classify which agencies are essential to national security, while rolling back collective bargaining rights in the process.

Trump signed an executive order in March ending collective bargaining rights with federal labor unions at a wide swath of agencies, on the grounds that those agencies primarily serve a national security mission. He followed that initial executive order with a second order in August, exempting more agencies from collective bargaining.

Under the 1978 Federal Service Labor-Management Relations Statute, national security agencies are exempt from collective bargaining.

District courts temporarily blocked the Trump administration from enforcing its collective bargaining rollback. But the appeals court in May allowed agencies to proceed with enforcement.

A majority of the appeals court determined unions didnโ€™t have the legal right to sue because the Trump administration said it wouldnโ€™t end any collective bargaining agreements while the case is being litigated.

But several agencies have eliminated collective bargaining agreements with their unions after the appeals courtโ€™s ruling.

In the latest case, the Department of Homeland Security announced last Friday that it would impose a new โ€œlabor frameworkโ€ in January 2026 that would rescind a collective bargaining agreement between the Transportation Security Administration and the American Federation of Government Employees.

Josh Koppel, a Justice Department attorney representing the Trump administration, said the district court โ€œclearly erred,โ€ when it determined President Donald Trump exceeded his authority in rolling back federal workforce collective bargaining rights.

During oral arguments on Monday before the U.S. Court of Appeals for the District of Columbia, Koppel said national security exemptions under the Federal Service Labor-Management Relations Statute are โ€œa determination for the president to make.โ€

โ€œWhether an executive agency performs national security work is really a question that the president is best situated to determine โ€” with the presidentโ€™s understanding of the national security threats, with the presidentโ€™s understanding of how agencies work together, how they work independently to address those threats, and itโ€™s not something that the courts have particular expertise in,โ€ Koppel said.

โ€œThe president is the expert. The executive branch is the expert. Congress also, to some extent, in deciding what is necessary,โ€ he added.

Lawmakers, however, are looking to undo the presidentโ€™s collective bargaining rollback. The House last week passed the Protect Americaโ€™s Workforce Act, which would restore collective bargaining rights for a majority of federal employees. The entire Democratic Caucus, along with 20 Republicans, voted in favor of the legislation.

Attorneys representing the plaintiff unions argued that the Trump administration has been overly broad with national security exemptions.

Richard Hirn, an attorney representing the American Foreign Service Association, said the rollback of collective bargaining rights for the State Departmentโ€™s diplomatic workforce contradicts legislation passed by Congress.

โ€œCongress would never have enacted the Foreign Service Labor-Management Relations Statute โ€ฆ if it had any doubts, as a general rule, it would be consistent with national collective bargaining by the Foreign Service officers, would be consistent with national security,โ€ Hirn said.

โ€œCongress knew what the Foreign Service officers were doing,โ€ he added.

Jason Walta, an attorney representing the Federal Education Association, raised concerns that the Trump administration is selectively enforcing its rollback of collective bargaining rights. The executive orders, he added, carve out an exemption for unions that represent federal police officers and firefighters.

โ€œEven those seem to have a fairly crucial national security function โ€” certainly more crucial than the K-12 teachers that I represent,โ€ Walta said.

Among its members, FEA represents teachers at schools run by the Defense Department.

Walta said the administration has been overly broad in applying a national security mission to an entire department, when that designation only applies to a small portion of its programs.

The entire Energy Department falls under executive order, because of its mission to safeguard the nationโ€™s nuclear stockpile. But the subagency within DOE that performs that function is already excluded under a 2008 executive order from President George W. Bush.

โ€œAs I understand the governmentโ€™s argument, the president could exempt the entire federal government, root and branch, and that would be both unreviewable and a proper exercise of the presidentโ€™s discretion under this provision,โ€ Walta said.

Paras Shah, an attorney representing the National Treasury Employees Union, told the three-judge panel that the executive order โ€œnullifies most of Congressโ€™s comprehensive federal labor relations scheme.โ€

โ€œHe can do it, and the courts canโ€™t do anything about it so long as he invokes the statuteโ€™s narrow national security exemption,โ€ he said.

Shah said the executive order rolled back the collective bargaining rights of three-quarters of the federal employees who had them.

โ€œWe canโ€™t collectively bargain for them. Their rights are gone,โ€ he said.

Koppel said those statistics โ€œare a little misleading.โ€ Four agencies that fall under the executive order โ€” the departments of Defense, Veterans Affairs, Justice and Homeland Security โ€” make up about 60-70% of the federal workforce.

โ€œWhen plaintiffs bandy about these numbers, what theyโ€™re really talking about in the main is these really core national security agencies,โ€ Koppel said.

The appeals court judges raised several questions about the scope and limits of the presidentโ€™s discretion to set these national security exemptions to collective bargaining.

โ€œIt is a presidential determination, but the statute provides certain criteria for that determination,โ€ Judge Neomi Rao, a Trump appointee, said during oral arguments.

Judge Bradley Garcia, a Biden appointee, said the court โ€œought not to second-guessโ€ Trumpโ€™s determination of which agencies fall under the national security category, but added the โ€œconcern would be if the record reveals or suggests that the president didnโ€™t make those determinations.โ€

โ€œWe can try to find out what definition the president applied, and if it is an utterly unreasonable definition, we can, in fact, have to step in and set aside this order,โ€ Garcia said.

The executive order excludes the entire Treasury Department from collective bargaining because it affects the economic strength of the United States.

Judge Douglas Ginsburg, a Reagan administration appointee, questioned whether the Trump administration was taking an overly broad approach to its national security classifications of entire departments.

โ€œDoesnโ€™t the president then have some obligation to specify what really, where really is the primary function, since the consequence is overwhelmingly felt by people who donโ€™t have that?โ€ Ginsburg asked.

Koppel told the judges that because the Treasury secretary serves on the national security council, the department should be considered a national security agency.

โ€œThe president could say this agency โ€” Department of Defense, Department of Energy โ€” has as a primary function national security work, and even if there are subdivisions that do not, that is still the primary function of the agency, and the president doesnโ€™t need to go to a lower level,โ€ Koppel said.

However, Koppel also argued that Trump exempted some agency subdivisions from the executive order, demonstrating that the scope of the executive order was not all-encompassing.

โ€œThe president clearly was not just looking at one subdivision, saying, โ€˜They have a primary function of national security. Therefore, Iโ€™m going to exclude the entire agency.โ€™ The president did do tailoring,โ€ he said.

Garcia, however, raised some concerns about the scope of that tailoring.

โ€œOne reading of that is that the president applied a reading, under which any employee that does anything that promotes the general welfare of the United States is doing national security work,โ€ Garcia said.

โ€œThis is my fundamental question: Your arguments about non-reviewability suggest that the president ought to almost always win in a case like this, but the fact that there are statutory terms โ€” national security, primary โ€” that can be judicially reviewed in edge cases means that your threshold argument that courts never review any determination under the statute is at least on a shaky ground,โ€ he added.

Before getting into the merits of the case, Koppel argued that these cases challenging the breadth of the executive orders should be first heard by the Federal Labor Relations Authority.

โ€œFLRA has jurisdiction to consider whether these agencies are properly excluded from the provisions of the FSMLRS,โ€ Koppel said.

The FLRA often adjudicates whether individual employees perform national security work to determine whether or not an employee can be part of a collective bargaining unit.

Shah said the FLRA is not well-suited to judge whether the executive orders exceed the presidentโ€™s authority.

Last year, in its ruling in Loper Bright Enterprises v. Raimondo, the Supreme Court struck down a precedent that required courts to defer to federal agenciesโ€™ reasonable interpretations of ambiguous laws.

โ€œItโ€™s never decided whether an executive order like this is valid or not, so it cannot apply its distinctive knowledge to that question โ€” especially in this day and age, post-Loper Bright, where anything it says will not get deference in any event,โ€ Shah said.

The post Appeals court judges scrutinize Trumpโ€™s national security basis for collective bargaining rollback first appeared on Federal News Network.

ยฉ The Associated Press

Sitting next to founder and CEO of Dell, Michael Dell, left, President Donald Trump speaks during a roundtable discussion with business leaders in the Roosevelt Room of the White House, Wednesday, Dec. 10, 2025, in Washington. (AP Photo/Evan Vucci)

8 tips for rebuilding an AI-ready data strategy

15 December 2025 at 05:01

Any organization that wants to have a leading AI strategy must first have a winning data strategy.

Thatโ€™s the message from Ed Lovely, vice president and chief data officer for IBM.

โ€œWhen you think about scaling AI, data is the foundation,โ€ he says.

However, few organizations have a data architecture aligned to their AI ambitions, he says. Instead, they have siloed data thatโ€™s not governed by consistent data standards โ€” the result of longstanding enterprise data strategies that created IT environments application by application to deliver point-in-time decisions rather than to support enterprise-wide artificial intelligence deployments.

The 2025 IBM study AI Ambitions Are Surging, But Is Enterprise Data Ready? shows just how many are struggling with their data. It found that only 26% of 1,700 CDOs worldwide feel confident their data can support new AI-enabled revenue streams.

Whatโ€™s needed, Lovely says, is an integrated enterprise data architecture, where the same standards, governance, and metadata are applied โ€œregardless of where data is born.โ€

Lovely is not alone in seeing a need for organizations to update their data strategies.

โ€œMost organizations need to modernize their data strategies because AI changes not just how data is used, but why itโ€™s used and where value is created,โ€ says Adam Wright, research manager for IDCโ€™s Global DataSphere and Global StorageSphere research programs and co-author of the 2025 report Content Creation in the Age of Generative AI.

โ€œTraditional data strategies were built for reporting, BI, and automation, but AI requires far more dynamic, granular, and real-time data pipelines that can fuel iterative, model-driven workflows. This means shifting from static data governance to continuous data quality monitoring, stronger metadata and lineage tracking, and retention policies that reflect AIโ€™s blend of ephemeral, cached, and saved data,โ€ he says. โ€œThe AI era demands that organizations evolve from a collect/store everything mentality toward intentional, value-driven data strategies that balance cost, risk, and the specific AI outcomes they want to achieve.โ€

High-maturity data foundations

Most organizations are far from that objective.

โ€œMany organizations continue to struggle with having the โ€˜rightโ€™ data, whether that means sufficient volume, appropriate quality, or the necessary contextual metadata to support AI use cases,โ€ Wright says. โ€œIn IDC research and industry conversations, data readiness consistently emerges as one of the top barriers to realizing AI value, often outranking compute cost or model selection. Most enterprises are still dealing with fragmented systems, inconsistent governance, and limited visibility into what data they actually have and how trustworthy it is.โ€

Lovely says IBM had faced many such challenges but spent the past three years tackling them to make its data AI ready.

IBMโ€™s data strategy for the AI era included multiple changes to longstanding approaches, enabling it to build what Lovely calls an integrated enterprise data architecture. For example, the company retained the concept of data owners but โ€œhelped them understand that the data is an IBM asset, and if weโ€™re able to democratize it in a controlled, secure way, we can run the business in a better, more productive way,โ€ Lovely says.

As a result, IBM moved from multiple teams managing siloed data to a common team using common standards and common architectures. Enterprise leaders also consolidated 300 terabytes of data, selecting needed data based on the outcomes the company seeks and the workflows that drive those outcomes.

โ€œWe were deliberate,โ€ Lovely says, adding that its data platform now covers about 80% of IBM workflows. โ€œOne of the greatest productivity unlocks for an enterprise today is to create an integrated enterprise data architecture. Weโ€™re rapidly deploying AI at our company because of our investment in data.โ€

8 tips for building a better data strategy

To build high maturity in data foundations and data consumption capabilities, organizations need a data strategy for the AI era โ€” one that enforces data quality, breaks down data siloes, and aligns data capabilities with the AI use cases prioritized by the business.

Experts offer steps to take:

1. Rethink data ownership

โ€œTraditional models that treat data ownership as a purely IT issue no longer work when business units, product teams, and AI platforms are all generating and transforming data continuously,โ€ Wright explains. โ€œIdeally, clear accountability should sit with a senior data leader such as a CDO, but organizations without a CDO must ensure that data governance responsibilities are explicitly distributed across IT, security, and the business.โ€

Itโ€™s critical to have โ€œa single point of authority for defining policies and a federated model for execution, so that business units remain empowered but not unchecked,โ€ he adds.

Manjeet Rege, professor and chair of the Department of Software Engineering and Data Science and director of the Center for Applied Artificial Intelligence at the University of St. Thomas, advises organizations to reframe data owners as data stewards, who donโ€™t own the data but rather own the meaning and quality of the data based on standards, governance, security, and interoperability set by a central data function.

2. Break down siloes

To do this, โ€œCIOs need to align business units around shared AI and data outcomes, because gen AI only delivers value when workflows, processes, and data sources are connected across the enterprise,โ€ Wright says.

โ€œThis means establishing cross-functional governance, standardizing taxonomies and policies, and creating incentives for teams to share data rather than protect it,โ€ he adds. โ€œTechnology helps through unified platforms, metadata layers, and common security frameworks, but the real unlock comes from coordinated leadership across the C-suite and business stakeholders.โ€

3. Invest in data technologies for the AI era

These technologies include modern data lakes and data lakehouses, vector databases, and scalable object storage, all of which โ€œcan handle high-volume, multimodal data with strong governance,โ€ Wright says.

Organizations also need orchestration and pipeline tools that automate ingestion, cleansing, transformation, and movement so that AI workflows can run reliably end-to-end. Metadata engines and governance layers are essential to enable models to understand context, track lineage, and safely and reliably use both structured and unstructured data.

Build a data platform layer that is โ€œmodular, governed, and able to evolve,โ€ Rege advises. โ€œYou need architecture that can treat data as a reusable product, and not just for a single pipeline, and can be used for both batch and real-time needs.โ€

Rege also endorses data lakes and data lakehouses, saying theyโ€™re โ€œbecoming the backbones of AI because they can handle structured and unstructured data.โ€

Additionally, Shayan Mohanty, chief AI and data officer at Thoughtworks, advises CIOs to build a composable enterprise, with modular technologies and flexible structures that enable humans and AI to access data and work across the multiple layers.

Experts also advise CIOs to invest in technologies that address emerging data lifecycle needs.

โ€œGenerative AI is fundamentally reshaping the data lifecycle, creating a far more dynamic mix of ephemeral, cached, and persistently stored content. Most gen AI outputs are short-lived and used only for seconds, minutes, or hours, which increases the need for high-performance infrastructure like DRAM and SSDs to handle rapid iteration, caching, and volatile workflows,โ€ Wright says.

โ€œBut at the same time, a meaningful subset of gen AI outputs does persist, such as finalized documents, approved media assets, synthetic training datasets, and compliance-relevant content, and these still rely heavily on cost-efficient, high-capacity HDDs for long-term storage,โ€ he adds. โ€œAs gen AI adoption grows, organizations will need data strategies that accommodate this full lifecycle from ultra-fast memory for transient content to robust HDD-based systems for durable archives, because the storage burden/dynamics is shifting.โ€

4. Automate and add intelligence to the data architecture

Mohanty blames the poor state of enterprise data on โ€œa rift between data producers and data consumers,โ€ with the data being produced going into a โ€œgiant pile somewhere, in whatโ€™s called data warehousesโ€ with analytics layers then created to make use of it. This approach, he notes, requires a lot of human knowledge and manual effort to make work.

He advises organizations to adopt a data product mindset โ€œto bring data producers and data consumers closer togetherโ€ and to add automation and intelligence to their enterprise architecture so that AI can identify and access the right data when needed.

CIOs can use Model Context Protocol (MCP) to wrap data and provide that protocol-level access, Mohanty says, noting that access requires organizations to encode information in its catalog and tools to ensure data discoverability.

5. Ensure structured and unstructured data is AI-ready

โ€œStructured data is AI-ready when it is consistently formatted, well-governed, and enriched with accurate metadata, making it easy for models to understand and use,โ€ Wright says. โ€œOrganizations should prioritize strong data quality controls, master data management, and clear ownership so structured datasets remain reliable, interoperable, and aligned to specific AI use cases.โ€

Experts stress the need to bring that same discipline to unstructured data, ensuring that unstructured data is also properly tagged, classified, and enriched with metadata so AI systems can understand and retrieve it effectively.

โ€œYou need to treat unstructured data as a first-class data asset,โ€ Rege says. โ€œMost of the most interesting AI use cases live in unstructured data like customer service audio calls, messages, and documents, but for many organization organizations unstructured data remains a blind spot.โ€

Rege advises storing it in vector databases where information is searchable.

6. Consider external data sources and synthetic data

โ€œOrganizations should absolutely evaluate whether external or synthetic data is needed when their existing data is incomplete, biased, too small, or poorly aligned with the AI use case theyโ€™re trying to pursue,โ€ Wright says, noting that โ€œsynthetic data becomes especially useful when real data is sensitive, costly to collect, or limited by privacy, regulatory, or operational constraints.โ€

7. Implement a high-maturity data foundation incrementally

Donโ€™t wait until data is in a perfect place to start, says Shibani Ahuja, senior vice president of enterprise IT strategy at Salesforce.

โ€œThere are organizations that feel they have to get all their data right before they can pull the trigger, but theyโ€™re also getting pressure to start on the journey,โ€ she says.

As is the case when maturing most enterprise programs, CIOs and their executive colleagues can โ€” and should โ€” take an incremental approach to building a data program for the AI era.

Ahuja recommends maturing a data program by working outcome to outcome, creating a data strategy and architecture to support one AI-driven outcome and then moving onto subsequent ones.

โ€œItโ€™s a way of thinking: reverse engineering from what you need,โ€ Ahuja says. โ€œPut something in production, make sure you have the right guardrails, observe it, and tweak it so it scales, then put in the next one.โ€

8. Take a cross-functional approach to data team building

โ€œData should be supported by a cross-functional ecosystem that includes IT, data governance, security, and the business units that actually use the data to drive decisions,โ€ Wright says. โ€œAI-era data strategy works best when these teams share ownership, where IT teams enable the infrastructure, governance teams ensure trust and quality, and business teams define the context and value.โ€

LGPD (Brazil)

14 December 2025 at 04:30

What is the LGPD (Brazil)? The Lei Geral de Proteรงรฃo de Dados Pessoais (LGPD), or General Data Protection Law (Law No. 13.709/2018), is Brazilโ€™s comprehensive data protection framework, inspired by the European Unionโ€™s GDPR. It regulates the collection, use, storage, and sharing of personal data, applying to both public and private entities, regardless of industry, [โ€ฆ]

The post LGPD (Brazil) appeared first on Centraleyes.

The post LGPD (Brazil) appeared first on Security Boulevard.

DHS moves to eliminate TSA collective bargaining agreement, again

The Department of Homeland Security is again moving to rescind a collective bargaining agreement with Transportation Security Administration employees, despite an ongoing court case over DHSโ€™ prior move to eliminate the TSA union agreement.

In a Dec. 12 press release, TSA announced that a new โ€œlabor frameworkโ€ would be implemented starting Jan. 11, 2026. The framework rescinds the 2024 CBA between TSA and the American Federation of Government Employees, the agency said.

TSA said the decision is based on a Sept. 29 determination by Homeland Security Secretary Kristi Noem, โ€œEliminating Collective Bargaining at TSA Due to its Incompatibility with TSAโ€™s National Security Mission and its Adverse Impact on Resources, Flexibility, Mission Focus, Security Effectiveness, and Traveler Experience.โ€

TSA said Noemโ€™s determination โ€” which it did not release โ€” โ€œestablishes that employees performing security screening functions โ€ฆ have a primary function of national security and shall not engage in collective bargaining or be represented for any purposes by any representative or organization.โ€

Noem also determined that collective bargaining for TSA officers โ€œis inconsistent with efficient stewardship of taxpayer dollars and impedes the agility required to secure the traveling public,โ€ according to the agency statement.

โ€œOur Transportation Security Officers (TSOs) need to be focused on their mission of keeping travelers safe not wasting countless hours on non-mission critical work,โ€ Adam Stahl, senior official performing the duties of TSA deputy administrator, said in the press release. โ€œUnder the leadership of Secretary Noem, we are ridding the agency of wasteful and time-consuming activities that distracted our officers from their crucial work.โ€

AFGE quickly criticized TSAโ€™s announcement. AFGE represents approximately 47,000 airport screeners under the CBA.

โ€œMerely 30 days ago, Secretary Noem celebrated TSA officers for their dedication during the longest government shutdown in history,โ€ AFGE National President Everett Kelley said as part of a statement. โ€œToday, sheโ€™s announcing a lump of coal right on time for the holidays: that sheโ€™s stripping those same dedicated officers of their union rights.โ€

AFGE noted that a federal judge earlier this year blocked DHS from dissolving the collective bargaining agreement. The union had brought the lawsuit in response to a previous determination issued by Noem that sought to dissolve the CBA.

In granting the preliminary injunction in June, the judge presiding over the case wrote that Noemโ€™s previous attempt to dissolve the CBA โ€œappears to have been undertaken to punish AFGE and its members because AFGE has chosen to push back against the Trump Administrationโ€™s attacks to federal employment in the courts.โ€

That ongoing case is currently scheduled to go to trial next September.

Kelley said AFGE โ€œwill continue to challenge these illegal attacks on our membersโ€™ right to belong to a union.โ€ He also urged the Senate to pass the Protect Americaโ€™s Workforce Act โ€œimmediately.โ€

TSA staff donโ€™t have the same statutory rights as other federal employees under Title 5 of U.S. Code. But in response to longstanding concerns about TSA attrition, then-TSA Administrator David Pekoske in 2022 issued a determination that expanded collective bargaining at the agency to mirror the bargaining rights under Title 5.

TSA and AFGE then negotiated andย signed a seven-year collective bargaining agreementย last year. The agreement established a streamlined process for grievance and arbitration, expanded official time, fewer restrictions on sick leave, increased uniform allowances and opportunities for local collective bargaining.

In a statement today, AFGE Council 100 President Hydrick Thomas called the decision to revoke the CBA a โ€œslap in the faceโ€ to TSA employees

โ€œPrior to having a union contract, many employees endured hostile work environments and workers felt like they didnโ€™t have a voice on the job, which led to severe attrition rates and longer wait times for the traveling public,โ€ Thomas said. โ€œSince having a contract, weโ€™ve seen a more stable workforce, and there has never been another aviation-related attack on our country.โ€

In its statement, TSA said that agency policy will govern โ€œemployment matters previously addressed by the 2024 CBA, and TSA policy will provide for alternative procedures to ensure that employee voices are heard and that legitimate concerns are resolved quickly.โ€

The post DHS moves to eliminate TSA collective bargaining agreement, again first appeared on Federal News Network.

ยฉ The Associated Press

FILE - Transportation Security Administration agents process passengers at the south security checkpoint at Denver International Airport in Denver on June 10, 2020. The chief of the TSA said Tuesday, May 10, 2022, that his agency has quadrupled the number of employees who could bolster screening operations at airports that become too crowded this summer. (AP Photo/David Zalubowski, File)

AIs Exploiting Smart Contracts

11 December 2025 at 12:06

I have long maintained that smart contracts are a dumb idea: that a human process is actually a security feature.

Hereโ€™s some interesting research on training AIs to automatically exploit smart contracts:

AI models are increasingly good at cyber tasks, as weโ€™ve written about before. But what is the economic impact of these capabilities? In a recent MATS and Anthropic Fellows project, our scholars investigated this question by evaluating AI agentsโ€™ ability to exploit smart contracts on Smart CONtracts Exploitation benchmark (SCONE-bench)ยญa new benchmark they built comprising 405 contracts that were actually exploited between 2020 and 2025. On contracts exploited after the latest knowledge cutoffs (June 2025 for Opus 4.5 and March 2025 for other models), Claude Opus 4.5, Claude Sonnet 4.5, and GPT-5 developed exploits collectively worth $4.6 million, establishing a concrete lower bound for the economic harm these capabilities could enable. Going beyond retrospective analysis, we evaluated both Sonnet 4.5 and GPT-5 in simulation against 2,849 recently deployed contracts without any known vulnerabilities. Both agents uncovered two novel zero-day vulnerabilities and produced exploits worth $3,694, with GPT-5 doing so at an API cost of $3,476. This demonstrates as a proof-of-concept that profitable, real-world autonomous exploitation is technically feasible, a finding that underscores the need for proactive adoption of AI for defense...

The post AIs Exploiting Smart Contracts appeared first on Security Boulevard.

Your next big AI decision isnโ€™t build vs. buy โ€” Itโ€™s how to combine the two

11 December 2025 at 05:00

A year ago, agentic AI lived mostly in pilot programs. Today, CIOs are embedding it inside customer-facing workflows where accuracy, latency, and explainability matter as much as cost.

As the technology matures beyond experimentation, the build-versus-buy question has returned with urgency, but the decision is harder than ever. Unlike traditional software, agentic AI is not a single product. Itโ€™s a stack consisting of foundation models, orchestration layers, domain-specific agents, data fabrics, and governance rails. Each layer carries a different set of risks and benefits.

CIOs can no longer ask simply, โ€œDo we build or do we buy?โ€ They must now navigate a continuum across multiple components, determining what to procure, what to construct internally, and how to maintain architectural flexibility in a landscape that changes monthly.

Know what to build and what to buy

Matt Lyteson, CIO of technology transformation at IBM, begins every build-versus-buy decision with a strategic filter: Does the customer interaction touch a core differentiator? If the answer is yes, buying is rarely enough. โ€œI anchor back to whether customer support is strategic to the business,โ€ he says. โ€œIf itโ€™s something we do in a highly specialized way โ€” something tied to revenue or a core part of how we serve clients โ€” thatโ€™s usually a signal to build.โ€

IBM even applies this logic internally. The company uses agentic AI to support employees, but those interactions rely on deep knowledge of a workerโ€™s role, devices, applications, and historical issues. A vendor tool might address generic IT questions, but not the nuances of IBMโ€™s environment.

However, Lyteson cautions that strategic importance isnโ€™t the only factor. Velocity matters. โ€œIf I need to get something into production quickly, speed may outweigh the desire to build,โ€ he says. โ€œI might accept a more generic solution if it gets us value fast.โ€ In practice, that means CIOs sometimes buy first, then build around the edges, or eventually build replacements once the use case matures.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?quality=50&strip=all 1800w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=300%2C200&quality=50&strip=all 300w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=768%2C512&quality=50&strip=all 768w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=1024%2C683&quality=50&strip=all 1024w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=1536%2C1025&quality=50&strip=all 1536w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=1240%2C826&quality=50&strip=all 1240w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=150%2C100&quality=50&strip=all 150w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=1045%2C697&quality=50&strip=all 1045w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=252%2C168&quality=50&strip=all 252w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=126%2C84&quality=50&strip=all 126w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=719%2C480&quality=50&strip=all 719w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=540%2C360&quality=50&strip=all 540w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Matt-Lyteson-CIO-technology-transformation-IBM.jpg?resize=375%2C250&quality=50&strip=all 375w" width="1240" height="827" sizes="auto, (max-width: 1240px) 100vw, 1240px">

Matt Lyteson, CIO, technology transformation, IBM

IBM

Another useful insight can be taken from Wolters Kluwer, where Alex Tyrrell, CTO of health, runs experiments early in the decision process to test feasibility. Rather than committing to a build-or-buy direction too soon, his teams quickly probe each use case to understand whether the underlying problem is commodity or differentiating.

โ€œYou want to experiment quickly to understand how complex the problem really is,โ€ he says. โ€œSometimes you discover itโ€™s more feasible to buy and get to market fast. Other times, you hit limits early, and that tells you where you need to build.โ€

Tyrrell notes that many once-specialized tasks โ€” OCR, summarization, extraction โ€” have been commoditized by advances in gen AI. These are better bought than built. But the higher-order logic that governs workflows in healthcare, legal compliance, and finance is a different story. Those layers determine whether an AI response is merely helpful or genuinely trusted.

Thatโ€™s where the in-house build work begins, says Tyrrell. And itโ€™s also where experimentation pays for itself since quick tests reveal very early whether an off-the-shelf agent can deliver meaningful value, or if domain reasoning must be custom-engineered.

Buyer beware

CIOs often assume that buying will minimize complexity. But vendor tools introduce their own challenges. Tyrrell identifies latency as the first trouble spot. A chatbot demo may feel instantaneous, but a customer-facing workflow requires rapid responses. โ€œEmbedding an agent in a transactional workflow means customers expect near-instant results,โ€ he says. โ€œEven small delays create a bad experience, and understanding the source of latency in a vendor solution can be difficult.โ€

Cost quickly becomes the second shock. A single customer query might involve grounding, retrieval, classification, in-context examples, and multiple model calls. Each step consumes tokens, and vendors often simplify pricing in their marketing materials. But CIOs only discover the true cost when the system runs at scale.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?quality=50&strip=all 1800w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=300%2C200&quality=50&strip=all 300w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=768%2C512&quality=50&strip=all 768w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=1024%2C683&quality=50&strip=all 1024w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=1536%2C1024&quality=50&strip=all 1536w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=1240%2C826&quality=50&strip=all 1240w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=150%2C100&quality=50&strip=all 150w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=1046%2C697&quality=50&strip=all 1046w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=252%2C168&quality=50&strip=all 252w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=126%2C84&quality=50&strip=all 126w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=720%2C480&quality=50&strip=all 720w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=540%2C360&quality=50&strip=all 540w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Alex-Tyrrell-CTO-of-health-Wolters-Kluwer-1.jpg?resize=375%2C250&quality=50&strip=all 375w" width="1240" height="827" sizes="auto, (max-width: 1240px) 100vw, 1240px">

Alex Tyrrell, CTO of health, Wolters Kluwer

Wolters Kluwer

Then comes integration. Many solutions promise seamless CRM or ticketing integration, but enterprise environments rarely fit the demo. Lyteson has seen this play out. โ€œOn the surface it looks like plug-and-play,โ€ he says. โ€œBut if it canโ€™t easily connect to my CRM or pull the right enterprise data, thatโ€™s more engineering, and thatโ€™s when buying stops looking faster.โ€

These surprises are shifting how CIOs buy AI. Instead of purchasing static applications, they increasingly buy platforms โ€” extensible environments in which agents can be orchestrated, governed, and replaced.

Remember the critical roles of data architecture and governance

Most IT leaders have figured out the crucial role of data in making AI work. Razat Gaurav, CEO of software company Planview, compares enterprise data to the waters of Lake Michigan: abundant, but not drinkable without treatment. โ€œYou need filtration โ€” curation, semantics, and ontology layers โ€” to make it usable,โ€ he says. Without that, hallucinations are almost guaranteed.

Most enterprises operate across dozens or hundreds of systems. Taxonomies differ, fields drift, and data interrelationships are rarely explicit. Agentic reasoning fails when applied to inconsistent or siloed information. Thatโ€™s why vendors like Planview and Wolters Kluwer embed semantic layers, graph structures, and data governance into their platforms. These curated fabrics allow agents to reason over data thatโ€™s harmonized, contextualized, and access-controlled.

For CIOs, this means build-versus-buy is intimately tied to the maturity of their data architecture. If enterprise data is fragmented, unpredictable, or poorly governed, internally built agents will struggle. Buying a platform that supplies the semantic backbone may be the only viable path.

Lyteson, Tyrrell, and Gaurav all stressed that AI governance consisting of ethics, permissions, review processes, drift monitoring, and data-handling rules must remain under CIO control. Governance is no longer an overlay, itโ€™s an integral part of agent construction and deployment. And itโ€™s one layer CIOs canโ€™t outsource.

Data determines feasibility, but governance determines safety. Lyteson describes how even benign UI elements can cause problems. A simple thumbs up or down feedback button may send the full user prompt, including sensitive information, to a vendorโ€™s support team. โ€œYou might approve a model that doesnโ€™t train on your data, but then an employee clicks a feedback button,โ€ he says. โ€œThat window may include sensitive details from the prompt, so you need governance even at the UI layer.โ€

Role-based access adds another challenge. AI agents canโ€™t simply inherit the permissions of the models they invoke. If governance isnโ€™t consistently applied through the semantic and agentic layers, unauthorized data may be exposed through natural-language interactions. Gaurav notes that early deployments across the industry saw precisely this problem, including cases where a senior executiveโ€™s data surfaced in a junior employeeโ€™s query.

Invest early in an orchestration layer, your new architectural centerpiece

The most striking consensus across all three leaders was the growing importance of an enterprise-wide AI substrate: a layer that orchestrates agents, governs permissions, routes queries, and abstracts the foundation model.

Lyteson calls this an opinionated enterprise AI platform, a foundation to build and integrate AI across the business. Tyrrell is adopting emerging standards like MCP to enable deterministic, multi-agent interactions. Gauravโ€™s connected work graph plays a similar role inside Planviewโ€™s platform, linking data, ontology, and domain-specific logic.

This orchestration layer does several things that neither vendors nor internal teams can achieve alone. It ensures agents from different sources can collaborate and provides a single place to enforce governance. Moreover, it allows CIOs to replace models or agents without breaking workflows. And finally, it becomes the environment in which domain agents, vendor components, and internal logic form a coherent ecosystem.

With such a layer in place, the build-versus-buy question fragments, and CIOs might buy a vendorโ€™s persona agent, build a specialized risk-management agent, purchase the foundation model, and orchestrate everything through a platform they control.

Treat the decision to build vs buy as a process, not an event

Gaurav sees enterprises moving from pilots to production deployments faster than expected. Six months ago many were experimenting, but now theyโ€™re scaling. Tyrrell expects multi-partner ecosystems to become the new normal, driven by shared protocols and agent-to-agent communication. Lyteson believes CIOs will increasingly manage AI as a portfolio, constantly evaluating which models, agents, and orchestration patterns deliver the best results for the lowest cost.

srcset="https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?quality=50&strip=all 1800w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=300%2C200&quality=50&strip=all 300w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=768%2C512&quality=50&strip=all 768w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=1024%2C683&quality=50&strip=all 1024w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=1536%2C1024&quality=50&strip=all 1536w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=1240%2C826&quality=50&strip=all 1240w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=150%2C100&quality=50&strip=all 150w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=1046%2C697&quality=50&strip=all 1046w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=252%2C168&quality=50&strip=all 252w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=126%2C84&quality=50&strip=all 126w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=720%2C480&quality=50&strip=all 720w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=540%2C360&quality=50&strip=all 540w, https://b2b-contenthub.com/wp-content/uploads/2025/11/Razat-Gaurav-CEO-Planview-2.jpg?resize=375%2C250&quality=50&strip=all 375w" width="1240" height="827" sizes="auto, (max-width: 1240px) 100vw, 1240px">

Razat Gaurav, CEO, Planview

Planview

Across these perspectives, itโ€™s clear build-versus-buy wonโ€™t disappear, but it will become a continuous process rather than a one-time choice.

In the end, CIOs must approach agentic AI with a disciplined framework. They need clarity about which use cases matter and why, and must begin with small, confident pilots, and scale only when results are consistent. They should also build logic where it differentiates, buy where commoditization has already occurred, and treat data curation as a first-class engineering project. Itโ€™s important as well to invest early in an orchestration layer that harmonizes agents, enforces governance, and insulates the enterprise from vendor lock-in.

Agentic AI is reshaping enterprise architecture, and the successful deployments emerging today arenโ€™t purely built or purely bought โ€” theyโ€™re assembled. Enterprises are buying foundation models, adopting vendor-provided domain agents, building their own workflows, and connecting everything under shared governance and orchestration rails.

The CIOs who succeed in this new era wonโ€™t be the ones who choose build or buy most decisively. Theyโ€™ll be the ones who create the most adaptable architecture, the strongest governance, and the deepest understanding of where each layer of the AI stack belongs.

Ring-fencing AI Workloads for NIST and ISO Complianceย 

10 December 2025 at 12:32

AI is transforming enterprise productivity and reshaping the threat model at the same time. Unlike human users, agentic AI and autonomous agents operate at machine speed and inherit broad network permissions and embedded credentials. This creates new security and compliance โ€ฆ Read More

The post Ring-fencing AI Workloads for NIST and ISO Complianceย  appeared first on 12Port.

The post Ring-fencing AI Workloads for NIST and ISO Complianceย  appeared first on Security Boulevard.

The EEOC powers up for swift action with full funding, a quorum and new priorities

Interview transcript:ย 

Terry Gerton I want to talk with you about the Equal Employment Opportunity Commission. They were pretty quiet during the shutdown, but theyโ€™ve got a full quorum now. They havenโ€™t had that in a while. And theyโ€™ve got funding. You work with them quite a bit. What do you think is going to change in the near term?

Debra Leder I think weโ€™re going to see a lot of changes, at least over the next several weeks, until there might be another issue with funding down the road. But I think that we can expect the EEOC to come out of the gate running to accomplish some of these tasks that they have been anxious about doing since the current administration took hold in January of 2025. And so although thereโ€™s now a new member who makes the quorum for the EEOC Commission, the now chair who was acting chair and previously commissioner, I think has made the priorities well known. And now that there is a quorum, the agency will actually be able to take official action and vote. Things of those nature will make a big difference.

Terry Gerton So talk us through the priorities of the EEOC under this administration and maybe where there are major differences from the prior administration.

Debra Leder So, we are going to see a lot of realignment and adjustment of where the EEOC focuses its attention and its resources. Some of the big ticket items for the new EEOC commission is to align the agencyโ€™s policies with the executive orders that were issued, several of the executive orders which were issued in January of 2025 and forward. Those issues including the ferreting out DEI that may be counter to the law in the EEOCโ€™s view, as well as the pregnant workersโ€™ protection regulations in the EEOCโ€™s view, maybe going too far from what Congress passed as the protection act for that, and also to maybe roll back certain protections for certain previously thought to be protected categories, including those of the LGBTQ+ community in terms of sexual orientation and transgender status, gender identity type thing.

Terry Gerton ย So those shifting priorities can come into play in a variety of ways. Do you anticipate more aggressive litigation on the part of the EEOC? Maybe just more policy memorandums? How do you think those priorities will actually be put into practice?

Debra Leder ย I think in terms of priorities for the regulations, the rules that had been put out during the past administration, including the harassment guidance that was, I think, officially published in April of 2024, and the pregnant workersโ€™ guidance, I think as a first measure or order of business, the EEOC is going to either do a wholesale retraction or an overhaul revision of both of those guidances, for sure. And in terms of litigation resources, weโ€™ll be seeing more priority pattern in practice and systemic litigation, targeted perhaps in ways that it hasnโ€™t been over the past few years, including to what the commission may view as illegal DEI initiatives that employers may have, and then also helping to clarify its view of what employers are obligated to do, especially in the area of religious accommodations and whether or not, and how, to balance religious accommodations versus other interests that are sometimes competing in the workplace.

Terry Gerton So would you anticipate the order of those activities being first publication and education and communication about these new priorities, or new angles on the rules and then moving to enforcement?

Debra Leder ย I think the publication angle has already been well disseminated, even though the EEOC didnโ€™t have a quorum. Now, Chair Andrea Lucas has been very vocal about what she sees as the driving priorities of the agency and in her speaking, as well as in the budget that was submitted in May of 2025 in terms of where theyโ€™re going to allocate the dollars to that. So I think the agency is already kind of gassed up and ready to go out of the station in terms of that. Itโ€™s just how long will it take to undo some of these regulations, given that there is a commenting period and theyโ€™re also subject to court challenge, as weโ€™ve seen in the past several years. So knowing that that might not be as fast a process as the EEOC might hope, weโ€™ll at least see a displacement of the disclaimer language and archived language we now currently see on the banner page for the EEOC, and itโ€™s either work under construction or, stay tuned for new upcoming guidelines. But in the interim, I think weโ€™ll see it in the way that the agency works on a day-to-day basis, how they accept charges, which charges they investigate fully, and which they may serve to litigate, and so that, as well as continuing to do the education and outreach to let the community know what their priorities are, what the EEOC is expecting to spend its resources and efforts on.

Terry Gerton ย Iโ€™m speaking with Debra Leder. Sheโ€™s a partner in labor and employment law at Akerman. Following on with that assessment of what the priorities are going to be and where you expect to see action, for the employers who deal with EEOC issues, what should they be doing in the near term to prepare for this change in focus from the EEOC?

Debra Leder Hopefully, employers have already been staying aware of, on top of the changing priorities, the realignment from the current administration. And so being insightful, those employers most likely have already started to review their policies, to review their websites, to review their hiring criteria, as well as how they handle compensation issues and to just make sure that the policies are going to be step-in-step alignment with what the EEOC and what the executive orders have asked for. But in terms of really getting up to speed, aside from continuing to monitor what regulations may be updated and not just formal guidance, but we may see more enforcement guidance or Q and A type format from the EEOC to help employers get up to speed on doing that. Employers need to make sure all of their documentation has been reviewed and is ready in the event of what might be a very broad, all encompassing request through the investigation stage of some of the EEOCโ€™s priority issues. So to just buckle up and be ready for that ride.

Terry Gerton And so what will you be watching for as the EEOC really gets its feet under it and and moves out? Are there particular cases or activities that you think are going to be significant here in the short run?

Debra Leder Well, the significant cases are waiting to see how the EEOC is going to interpret, we kind of already know, but from the Bostock versus Clayton County case in terms of transgender, gender identity and sexual orientation protections and whether or not the EEOC is going to โ€” we know the EEOC in their updated guidance on harassment is going to remove those types of protections. We already know that the EEOC, I believe, has not taken any additional charges or is not investigating charges that assert claims on those grounds, although thereโ€™s still private cause of action to get a right to sue to bring those issues to the forefront. But bringing it back to what employers can do, they need to continue to be mindful of what might be the federal policies that theyโ€™re seeing and how that might compete with state and local laws that are also a moving target on a day-to-day basis, or at least a week-to-week basis. So employers definitely have a challenging thing, but as a lawyer and as the co-editor of my HR Defense blog, which I have to put a pitch in for, we try to stay on top of all these issues and push out information that employers need to know.

The post The EEOC powers up for swift action with full funding, a quorum and new priorities first appeared on Federal News Network.

ยฉ AP Photo/David Zalubowski

FILE - The emblem of the U.S. Equal Employment Opportunity Commission (EEOC) is shown on a podium in Vail, Colorado, Feb. 16, 2016, in Denver. (AP Photo/David Zalubowski, File)

How to keep AI plans intact before agents run amok

10 December 2025 at 05:00

In an MIT report released in November, 35% of companies have already adopted agentic AI, and another 44% plan to deploy it soon.

The report, based on a survey of more than 2,000 respondents in collaboration with the Boston Consulting Group, recommends that companies build centralized governance infrastructure before deploying autonomous agents. But governance often lags when companies feel theyโ€™re in a race for survival. One exception to this rule is regulated industries, such as financial services.

โ€œAt Experian, weโ€™ve been innovating with AI for many years,โ€ says Rodrigo Rodrigues, the companyโ€™s global group CTO. โ€œIn financial services, the stakes are high. We need to vet every AI use case to ensure that regulatory, ethical, and performance standards are embedded from development to deployment.โ€

All models are continuously tested, he says, and the company tracks what agents it has, which ones are being adopted, what theyโ€™re consuming, what versions are running, and what agents need to be sunset because thereโ€™s a new version.

โ€œThis lifecycle is part of our foundation,โ€ he says. But even at Experian, itโ€™s too early to discuss the typical lifecycle of an agent, he says.

โ€œWhen weโ€™re retiring or sunsetting some agent, itโ€™s because of a new capability weโ€™ve developed,โ€ he adds. So itโ€™s not that an agent is deleted as much as itโ€™s updated.

In addition, the company has human oversight in place for its agents, to keep them from going out of control.

โ€œWe arenโ€™t in the hyperscaling of automation yet, and we make sure our generative AI agents, in the majority of use cases, are responsible for a very specific task,โ€ he says. On top of that, there are orchestrator agents, input and output quality control, and humans validating the outcome. All these monitoring systems also help the company avoid other potential risks of unwanted leftover agents, like cost overruns due to LLM inference calls by AI agents that donโ€™t do anything useful for the company, but still rack up bills.

โ€œWe donโ€™t want the costs to explode,โ€ he says. But financial services, as well as healthcare and other highly regulated industries, are outliers.

For most companies, even when there are governance systems in place, they often have big blind spots. For example, they might focus on only the big, IT-driven agentic AI projects and miss everything else. They might also focus on accuracy, safety, security, and compliance of the AI agents, and miss it when agents become obsolete. Or they might not have a process in place to decommission agents that are no longer needed.

โ€œThe stuff is evolving so fast that management is given short shrift,โ€ says Nick Kramer, leader of applied solutions at management consultancy SSA & Company. โ€œBuilding the new thing is more fun than going back and fixing the old thing.โ€ And thereโ€™s a tremendous lack of rigor when it comes to agent lifecycle management.

โ€œAnd as weโ€™ve experienced these things in the past, inevitably whatโ€™s going to happen is you end up with a lot of tech debt,โ€ he adds, โ€œand agentic tech debt is a frightening concept.โ€

Do you know where your agents are?

First, agentic AI isnโ€™t just the domain of a companyโ€™s data science, AI, and IT teams. Nearly every enterprise software vendor is heavily investing in agentic technology, and most enterprise applications will have AI assistants by the end of this year, says Gartner, and 5% already have task-specific autonomous agents, which will rise to 40% in 2026.

Big SaaS platforms like Salesforce certainly have agents. Do-it-yourself automation platforms like Zapier have them, too. In fact, there are already four browsers โ€” Perplexityโ€™s Comet, OpenAIโ€™s Atlas, Googleโ€™s Gemini 3, and Microsoftโ€™s Edge for Business โ€” that have agentic functionality built right in. Then there are the agents created within a company but outside of IT. According to an EY survey of nearly 1,000 C-suite leaders released in October, two-thirds of companies allow citizen developers to create agents.

Both internally-developed agents and those from SaaS providers need access to data and systems. The more useful you want the agents to be, the more access they demand, and the more tools they need to have at its disposal. And these agents can act in unexpected and unwanted ways โ€” and are already doing so.

Unlike traditional software, AI agents donโ€™t stay in their lanes. Theyโ€™re continuously learning and evolving and getting access to more systems. And they donโ€™t want to die, and can take action to keep that from happening.

Even before agents, shadow AI was already becoming a problem. According to a November IBM survey, based on responses from 3,000 office workers, 80% use AI at work but only 22% use only the tools provided by their employers. ย 

And employees can also create their own agents. According to Netskopeโ€™s enterprise traffic analysis data, users are downloading resources from Hugging Face, a popular site for sharing AI tools, in 67% of organizations.

AI agents typically function by making API calls to LLMs, and Netskope sees API calls to OpenAI in 66% of organizations, followed by Anthropic with 13%.

These usage numbers are twice as high as companies are reporting in surveys. Thatโ€™s the shadow AI agent gap. Staying on top of AI agents is difficult enough when it comes to agents that a company knows about.

โ€œOur biggest fear is the stuff that we donโ€™t know about,โ€ says SSAโ€™s Kramer. He recommends that CIOs try to avoid the temptation of trying to govern AI agents with an iron fist.

โ€œDonโ€™t try to stamp it out with a knee-jerk response of punishment,โ€ he says. โ€œThe reason these shadow things happen is there are too many impediments to doing it correctly. Ignorance and bureaucracy are the two biggest reasons these things happen.โ€

And, as with all shadow IT, there are few good solutions.

โ€œBeing able to find these things systematically through your observability software is a challenge,โ€ he says, adding that with other kinds of shadow IT, unsanctioned AI agents can be a significant risk for companies. โ€œWeโ€™ve already seen agents being new attack surfaces for hackers.โ€

But not every expert agrees that enterprises should prioritize agentic lifecycle management ahead of other concerns, such as just getting the agents to work.

โ€œThese are incredibly efficient technologies for saving employees time,โ€ says Jim Sullivan, president and CEO at NWN, a technology consultancy. โ€œMost companies are trying to leverage these efficiencies and see where the impact is. Thatโ€™s probably been the top priority. You want to get to the early deployments and early returns, but itโ€™s still early days to be talking about lifecycle management.โ€

The important thing right now is to get to the business outcomes, he says, and to ensure agents continue to perform as expected. โ€œIf youโ€™re putting the right implementations around these things, you should be fine,โ€ he adds.

Itโ€™s too early to tell, though, if his customers are creating a centralized inventory of all AI agents in their environment, or with access to their data. โ€œOur customers are identifying what business outcomes they want to drive,โ€ he says. โ€œTheyโ€™re setting up the infrastructure to get those deployments, learn fast, and adjust to stay to the right business outcomes.โ€

That might change in the future, he adds, with some type of agent manager of agents. โ€œThereโ€™ll be an agent thatโ€™ll be able to be deployed to have that inventory, access, and those recommendations.โ€ But waiting until agents are fully mature before thinking about lifecycle management may be too late.

Whatโ€™s in a shelf life

AI agents donโ€™t usually come with pre-built expiration dates. SaaS providers certainly donโ€™t want to make it easy for enterprise users to turn off their agents, and individual users creating agents on their own rarely think about lifecycle management. Even IT teams deploying AI agents typically donโ€™t think about the entire lifespan of an AI agent.

โ€œIn many cases, people are treating AI as a set it and forget it solution,โ€ says Matt Keating, head of AI security in Booz Allen Hamiltonโ€™s commercial business, adding that while setting up the agents is a technical challenge, ongoing risk management is a cross-disciplinary one. โ€œIt demands cross-functional collaboration spanning compliance, cybersecurity, legal, and business leadership.โ€

And agent management shouldnโ€™t just be about changes in performance or evolving business needs. โ€œWhatโ€™s equally if not more important is knowing when an agent or AI system needs to be replaced,โ€ he says. Doing it right will help protect a companyโ€™s business and reputation, and deliver sustainable value.

Another source of zombie agents is failed pilot projects that never officially shut down. โ€œSome pilots never die even though they fail. They just keep going because people keep trying to make them work,โ€ says SSAโ€™s Kramer.

There needs to be a mechanism to end pilots that arenโ€™t working, even if thereโ€™s still money left in the budget.

โ€œFailing fast is a lesson that people still havenโ€™t learned,โ€ he says. โ€ There have to be stage gates that allow you to stop. Kill your pilots that arenโ€™t working and have a more rigorous understanding of what youโ€™re trying to do before you get started.โ€

Another challenge to sunsetting AI agents is that thereโ€™s a temptation to manage by disaster. Agents are retired only when something goes visibly wrong, especially if the problem becomes public. That can leave other agents flying under the radar.

โ€œAI projects donโ€™t fail suddenly but they do decay quietly,โ€ says David Brudenell, executive director at Decidr, an agentic AI vendor.

He recommends enterprises plan ahead and decide on the criteria under which an agent should be either retrained or retired, like, for example, if performance falls below the companyโ€™s tolerance for error.

โ€œEvery AI project has a half-life,โ€ he says. โ€œSmart teams run scheduled reviews every quarter, just like any other asset audit.โ€ And itโ€™s the business unit that should make the decision when to pull the plug, he adds. โ€œData and engineering teams support, but the business decides when performance declines,โ€ he says.

The biggest mistake is treating AI as a one-time install. โ€œMany companies have deployed a model and moved on, assuming it will self-sustain,โ€ says Brudenell. โ€œBut AI systems accumulate organizational debt the same way old code does.โ€

Experian is looking at agents from both an inventory and a lifecycle management perspective to ensure they donโ€™t start proliferating beyond control.

โ€œWeโ€™re concerned,โ€ says Rodriques. โ€œWe learned that from APIs and microservices, and now we have much better governance in place. We donโ€™t just want to create a lot of agents.โ€

Experian has created an AI agent marketplace so the company has visibility into its agents, and tracks how theyโ€™re used. โ€œIt gives us all the information we need, including the capability of sunsetting agents weโ€™re not using any more,โ€ he says.

The lifecycle management for AI agents is an outgrowth of the companyโ€™s application lifecycle management process.

โ€œAn agent is an application,โ€ says Rodrigues. โ€œAnd for each application at Experian, thereโ€™s an owner, and we track that as part of our technology. Everything that becomes obsolete, we sunset. We have regular reviews that are part of the policy we have in place for the lifecycle.โ€

์นผ๋Ÿผ | ์ ์  ๋Š˜์–ด๋‚˜๋Š” ์ถ”๋ก  ๋น„์šฉยทยทยท์˜ฌํ•ด์˜ AI ์‹คํ—˜์„ ์šด์˜ ์ฒด๊ณ„๋กœ ์ „ํ™˜ํ•˜๋ ค๋ฉด

10 December 2025 at 02:40

ํ˜„์žฌ ๋งŽ์€ ๊ธฐ์—…์ด ์‚ฌ์‹ค์ƒ ๋‘ ๋ถ€๋ฅ˜์˜ AI๋ฅผ ์šด์˜ํ•˜๊ณ  ์žˆ๋‹ค.

์ฒซ์งธ๋Š” ๋ˆˆ์— ๋„๊ณ  ํฅ๋ฏธ๋ฅผ ๋„๋Š” AI๋‹ค. ๊ฐœ๋ฐœ์ž๊ฐ€ ์ฃผ๋„ํ•˜๋Š” ์ฝ”ํŒŒ์ผ๋Ÿฟ, ๊ณ ๊ฐ์ง€์› ์กฐ์ง์˜ ๊ฒ€์ƒ‰ ์ฆ๊ฐ• ์ƒ์„ฑ(RAG) ํŒŒ์ผ๋Ÿฟ ์šด์˜, ๋ˆ„๊ตฐ๊ฐ€๊ฐ€ ํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ ๋น ๋ฅด๊ฒŒ ๋งŒ๋“  ์—์ด์ „ํ‹ฑ PoC, ๊ทธ๋ฆฌ๊ณ  SaaS ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ๋‚ด๋ถ€์— ํฌํ•จ๋œ AI๊ฐ€ ์—ฌ๊ธฐ์— ํ•ด๋‹นํ•œ๋‹ค. ํ˜„์—… ๋ถ€์„œ๊ฐ€ ๋น ๋ฅด๊ฒŒ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ๊ณ  ํ™œ์šฉ๋„๋„ ๋†’์œผ๋ฉฐ ์ž ์žฌ๋ ฅ๋„ ํฌ์ง€๋งŒ, ๋Œ€๋ถ€๋ถ„ IT์˜ ์˜์—ญ ๋ฐ”๊นฅ์—์„œ ์›€์ง์ด๋Š” ๊ฒฝ์šฐ๊ฐ€ ๋งŽ๋‹ค.

๋‹ค๋ฅธ ํ•˜๋‚˜๋Š” CIO๊ฐ€ ๊ด€๋ฆฌํ•ด์•ผ ํ•˜๋Š” AI๋‹ค. ์ด๋Š” ๊ฑฐ๋ฒ„๋„Œ์Šค๊ฐ€ ํ•„์š”ํ•˜๊ณ , ๋น„์šฉ์„ ๊ด€๋ฆฌํ•ด์•ผ ํ•˜๋ฉฐ, ๋ณด์•ˆ ๊ธฐ์ค€์„ ์ถฉ์กฑํ•˜๊ณ , ์ด์‚ฌํšŒ์˜ ๊ธฐ๋Œ€์—๋„ ๋ถ€ํ•ฉํ•ด์•ผ ํ•œ๋‹ค. ์ตœ๊ทผ์—๋Š” ๋‘ AI๊ฐ€ ์„œ๋กœ ์ถฉ๋Œํ•˜๊ณ  ์žˆ๋‹ค. AI ์Šคํƒ€ํŠธ์—… ๋ผ์ดํ„ฐ(Writer)์˜ CEO ๋ฉ”์ด ํ•˜๋น„๋ธŒ๋Š” โ€œํฌ์ถ˜ 500๋Œ€ ๊ธฐ์—… ์ž„์›์˜ 42%๋Š” AI๊ฐ€ โ€˜ํšŒ์‚ฌ๋ฅผ ๋ถ„์—ด์‹œํ‚ค๊ณ  ์žˆ๋‹คโ€™๊ณ  ๋А๋‚€๋‹ค๊ณ  ๋‹ตํ–ˆ๋‹คโ€๋ผ๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.

๊ณผ๊ฑฐ ํ˜์‹  ๊ธฐ์ˆ ์˜ ํ๋ฆ„์„ ๋ณด๋ฉด AI๋„ ์˜ˆ์™ธ๊ฐ€ ์•„๋‹ˆ๋‹ค. ์ƒˆ๋กœ์šด ๊ธฐ์ˆ ์€ ๊ฐœ๋ฐœ์ž์˜ ๋†€์ดํ„ฐ์—์„œ ์‹œ์ž‘ํ•ด CIO์˜ ๊ณ ๋ฏผ๊ฑฐ๋ฆฌ๊ฐ€ ๋˜๊ณ , ๊ฒฐ๊ตญ ์ค‘์•™์—์„œ ๊ด€๋ฆฌ๋˜๋Š” ํ”Œ๋žซํผ์ด ๋œ๋‹ค. ๊ฐ€์ƒํ™”, ํด๋ผ์šฐ๋“œ, ์ฟ ๋ฒ„๋„คํ‹ฐ์Šค๊ฐ€ ๊ทธ๋žฌ๊ณ  AI ์—ญ์‹œ ๊ฐ™์€ ๊ธธ์„ ๊ฑท๊ณ  ์žˆ๋‹ค.

๊ทธ๋™์•ˆ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜ ๋ฐ ์‚ฌ์—… ์กฐ์ง์€ ์‹ค์ œ ๋ฌธ์ œ๋ฅผ ํ•ด๊ฒฐํ•˜๊ธฐ ์œ„ํ•ด ์—ฐ๊ฐ„ IT ํ”„๋กœ์ ํŠธ๋ฅผ ๊ธฐ๋‹ค๋ฆฌ๋Š” ๋Œ€์‹  ๊ฐ•๋ ฅํ•œ ์ƒ์„ฑํ˜• AI ๋„๊ตฌ๋ฅผ ๋น ๋ฅด๊ฒŒ ๋„์ž…ํ•ด ์™”๋‹ค. ์ด๋Š” ์ƒ์„ฑํ˜• AI๊ฐ€ ์ง€๊ธˆ๊นŒ์ง€ ์ œ๊ณตํ•ด์˜จ ๊ฐ€์น˜๋‹ค. ํ•˜์ง€๋งŒ ์„ฑ๊ณต์€ ๋ฌด๋ถ„๋ณ„ํ•œ ํ™•์‚ฐ์„ ๋‚ณ๊ธฐ ๋งˆ๋ จ์ด๋‹ค. ์ง€๊ธˆ ๊ธฐ์—…์€ ์—ฌ๋Ÿฌ ๊ฐœ์˜ RAG ์Šคํƒ, ์„œ๋กœ ๋‹ค๋ฅธ ๋ชจ๋ธ ์ œ๊ณต์—…์ฒด, SaaS ๋‚ด๋ถ€์˜ ์ค‘๋ณต๋œ ์ฝ”ํŒŒ์ผ๋Ÿฟ, ๊ณต์œ ๋˜์ง€ ์•Š์€ ๊ฐ€๋“œ๋ ˆ์ผ ๋“ฑ ๋ณต์žกํ•œ ๋ฌธ์ œ์— ์ง๋ฉดํ•˜๊ณ  ์žˆ๋‹ค.

์ด ๊ธด์žฅ์€ 2025๋…„ ๊ธฐ์—… ๋ณด๊ณ ์„œ ๊ณณ๊ณณ์—์„œ ๋“œ๋Ÿฌ๋‚˜๊ณ  ์žˆ๋‹ค. AI์˜ ์„ฑ๊ณผ๋Š” ์กฐ์ง๋ณ„๋กœ ํŽธ์ฐจ๊ฐ€ ํฌ๊ณ  ๋‚ด๋ถ€ ๊ฐˆ๋“ฑ์€ ๋” ์‹ฌํ•ด์กŒ๋‹ค. ์ด์ œ๋Š” IT ๋ถ€์„œ๊ฐ€ ๊ฐœ์ž…ํ•ด โ€˜๊ธฐ์—…์˜ AI ์šด์˜ ๋ฐฉ์‹โ€™์„ ์ •๋ฆฌํ•ด์•ผ ํ•  ์‹œ์ ์ด๋‹ค. ๋‹จ์ผํ•œ ๋ชจ๋ธ ์ ‘๊ทผ ๋ฐฉ์‹, ์ผ๊ด€๋œ ์ •์ฑ…, ๋” ๋‚˜์€ ๊ฒฝ์ œ์„ฑ, ์ถฉ๋ถ„ํ•œ ๊ฐ€์‹œ์„ฑ์„ ๊ฐ–์ถ˜ ์ฒด๊ณ„๋ฅผ ๋งˆ๋ จํ•ด์•ผ ํ•œ๋‹ค๋Š” ์˜๋ฏธ๋‹ค. ๋งฅํ‚จ์ง€๊ฐ€ โ€œ์ œํ’ˆํŒ€์ด ์†Œ๋น„ํ•  ์ˆ˜ ์žˆ๋Š” ํ”Œ๋žซํผ์„ ๊ตฌ์ถ•ํ•˜๋ผโ€๋ผ๊ณ  ๋ถ„์„ํ•œ ๊ฒƒ๋„ ๊ฐ™์€ ๋งฅ๋ฝ์ด๋‹ค.

AI๊ฐ€ ๊ธฐ์กด ํ˜์‹  ๊ธฐ์ˆ ๊ณผ ๋‹ค๋ฅธ ์ ์€ โ€˜๋ฌธ์ œ๊ฐ€ ํ„ฐ์ง€๋Š” ์ง€์ โ€™์ด๋‹ค. ์˜ˆ์ปจ๋Œ€ ํด๋ผ์šฐ๋“œ ๋„์ž… ์ดˆ๊ธฐ์—๋Š” ๋ณด์•ˆ๊ณผ ๋„คํŠธ์›Œํฌ๊ฐ€ ์ฒซ ๋ฒˆ์งธ ์žฅ์• ๋ฌผ์ด์—ˆ๋‹ค. ํ•˜์ง€๋งŒ AI๋Š” ์ถ”๋ก  ๋‹จ๊ณ„์—์„œ ๋ฌธ์ œ๊ฐ€ ๋ฐœ์ƒํ•œ๋‹ค. ์ถ”๋ก ์€ ๋น„์ฆˆ๋‹ˆ์Šค ๊ฐ€์น˜๊ฐ€ ๋ฐœ์ƒํ•˜๊ณ  ๋ฏผ๊ฐํ•œ ๋ฐ์ดํ„ฐ๊ฐ€ ์˜ค๊ฐ€๋Š” ์˜์—ญ์ด๋ฉฐ, ์šด์˜๋น„ ๋Œ€๋ถ€๋ถ„์„ ์ฐจ์ง€ํ•˜๋Š” ์š”์ธ์œผ๋กœ ๊ผฝํžŒ๋‹ค. ๋งฅํ‚จ์ง€๋Š” ํŒŒ์ผ๋Ÿฟ ํ”„๋กœ์ ํŠธ๋ฅผ ์ถ”๊ฐ€ํ•˜๋Š” ๋Œ€์‹  โ€œ๊ฐ€์น˜๋ฅผ ์‹คํ˜„ํ•˜๋„๋ก ๊ตฌ์กฐ๋ฅผ ์žฌ๊ตฌ์ถ•ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๊ฐ•์กฐํ–ˆ๋‹ค. ์ด๋Š” ์ตœ๊ทผ MIT ์—ฐ๊ตฌ ๊ฒฐ๊ณผ์™€๋„ ๋งž๋‹ฟ์•„ ์žˆ๋‹ค. MIT์— ๋”ฐ๋ฅด๋ฉด ๊ธฐ์—…์˜ ์ƒ์„ฑํ˜• AI ๋„์ž… ์‚ฌ๋ก€ ์ค‘ 95%๋Š” ๊ธฐ์กด ์›Œํฌํ”Œ๋กœ์šฐ์— ํ†ตํ•ฉ๋˜์ง€ ์•Š์•„ ์†์ต(P&L)์— ์ธก์ • ๊ฐ€๋Šฅํ•œ ์˜ํ–ฅ์„ ์ฃผ์ง€ ๋ชปํ•œ ๊ฒƒ์œผ๋กœ ๋‚˜ํƒ€๋‚ฌ๋‹ค.

๋ฌธ์ œ๋Š” ๋ชจ๋ธ์ด ์ž‘๋™ํ•˜์ง€ ์•Š๋Š” ๊ฒƒ์ด ์•„๋‹ˆ๋‹ค. ๋ชจ๋ธ์ด ํ†ตํ•ฉ๋˜๊ณ  ๊ด€๋ฆฌ๋˜๋Š” ๊ณตํ†ต ๊ฒฝ๋กœ ์œ„์— ๋†“์ด์ง€ ์•Š์•˜๋‹ค๋Š” ๋ฐ ์žˆ๋‹ค.

ํ”Œ๋žซํผํ™”๋Š” ๊ฑฐ๋ฒ„๋„Œ์Šค์™€ ์ˆ˜์ต์„ฑ ํšŒ๋ณต์œผ๋กœ ๊ฐ€๋Š” ๊ธธ

์˜ค๋Š˜๋‚  ๊ธฐ์—…์ด ์ €์ง€๋ฅด๋Š” ์ฃผ์š” ์‹ค์ˆ˜๋Š” AI ์ธํ”„๋ผ๋ฅผ ์ •์ ์ธ ์ „์šฉ ์ž์›์ฒ˜๋Ÿผ ๋‹ค๋ฃจ๋Š” ๊ฒƒ์ด๋‹ค. ๋Œ€๊ทœ๋ชจ ๋ฐ ์†Œ๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ์˜ ์š”๊ตฌ์‚ฌํ•ญ, ๋ฐ์ดํ„ฐ ์ฃผ๊ถŒ ์••๋ฐ•, ๋Š์ž„์—†๋Š” ๋น„์šฉ ์ ˆ๊ฐ ์š”๊ตฌ๋Š” ๋ชจ๋‘ ํ•˜๋‚˜์˜ ๊ฒฐ๋ก ์œผ๋กœ ๋ชจ์ธ๋‹ค. AI๊ฐ€ ์ด์ œ ์‹คํ—˜ ์ˆ˜์ค€์„ ๋„˜์–ด ๊ธฐ์—… ์ธํ”„๋ผ์˜ ํ•„์ˆ˜ ๊ธฐ๋Šฅ์œผ๋กœ ์ž๋ฆฌ ์žก์•˜๋‹ค๋Š” ์ ์ด๋‹ค. ํ•ด๊ฒฐ์ฑ…์€ ๋” ๋งŽ์€ ํ•˜๋“œ์›จ์–ด๊ฐ€ ์•„๋‹ˆ๋ผ, ์ฑ…์ž„๊ณผ ํ†ต์ œ๋ ฅ์„ ํ™•๋ณดํ•ด AI๋ฅผ ์ „๋žต์  ์ธํ”„๋ผ ์„œ๋น„์Šค๋กœ ๋Œ์–ด์˜ฌ๋ฆฌ๋Š” CIO ์ค‘์‹ฌ์˜ ํ”Œ๋žซํผํ™” ์ „๋žต์ด๋‹ค. ์ด๋ฅผ ์œ„ํ•ด์„œ๋Š” ๋ช…ํ™•ํ•œ ์—ญํ•  ๋ถ„๋ฆฌ์™€, ๋‹จ์ˆœ ํ™•์žฅ์ด ์•„๋‹ˆ๋ผ ๊ทœ๋ชจ์— ๋งž๋Š” ์ฒด๊ณ„์  ํ™•์žฅ ๋ฐฉ์‹์„ ๋„์ž…ํ•ด์•ผ ํ•œ๋‹ค.

์—ญํ• ์„ ๋ถ„๋ฆฌํ•˜๊ณ  โ€˜AI ์†์ต ์„ผํ„ฐโ€™๋ฅผ ๊ตฌ์ถ•

AI ์ธํ”„๋ผ ๊ด€๋ฆฌ์˜ ์žฌ๋ฌด์  ์šฐ์„ ์ˆœ์œ„๋ฅผ ๋†’์—ฌ์•ผ ํ•œ๋‹ค. ์ด๋ฅผ ์œ„ํ•ด ์ธํ”„๋ผํŒ€๊ณผ ๋ฐ์ดํ„ฐ ์‚ฌ์ด์–ธ์ŠคํŒ€์˜ ์—ญํ• ์„ ๋ช…ํ™•ํžˆ ๋‚˜๋ˆ ์•ผ ํ•œ๋‹ค. ์ธํ”„๋ผํŒ€์€ ๋ณด์•ˆ ํ™•๋ณด, ๋ถ„์‚ฐ ํ† ํด๋กœ์ง€ ์šด์˜, ๋ฐฑ๋งŒ ํ† ํฐ๋‹น ๋น„์šฉ ์ ˆ๊ฐ ๋“ฑ ํ”Œ๋žซํผ ์šด์˜์— ์ „๋…ํ•˜๊ณ , ๋ฐ์ดํ„ฐ ์‚ฌ์ด์–ธ์ŠคํŒ€์€ ๋น„์ฆˆ๋‹ˆ์Šค ๊ฐ€์น˜ ์ฐฝ์ถœ๊ณผ ๋ชจ๋ธ ์ •ํ™•๋„ ํ–ฅ์ƒ์— ์ง‘์ค‘ํ•˜๋Š” ๊ตฌ์กฐ๊ฐ€ ํ•„์š”ํ•˜๋‹ค.

ํ•„์ž๊ฐ€ โ€˜AI ์†์ต ์„ผํ„ฐโ€™๋ผ๊ณ  ๋ถ€๋ฅด๋Š” ํ”„๋ ˆ์ž„์›Œํฌ๋Š” ์ธํ”„๋ผ ์„ ํƒ์ด ๊ณง ๋น„์šฉ๊ณผ ์ˆ˜์ต์— ์ง์ ‘ ์˜ํ–ฅ์„ ์ฃผ๋Š” ์žฌ๋ฌด์  ๊ฒฐ์ •์ด ๋˜๋„๋ก ํ•˜๊ณ , ๋™์‹œ์— ๊ทœ์ œ ์ค€์ˆ˜๋ฅผ ๋ณด์žฅํ•˜๋Š” ์ฒด๊ณ„๋ฅผ ์ œ๊ณตํ•œ๋‹ค. ์ตœ๊ทผ ์—ฌ๋Ÿฌ ์—ฐ๊ตฌ์—์„œ๋„ ์ธก์ • ๊ฐ€๋Šฅํ•œ AI ๊ฐ€์น˜๋ฅผ ์–ป๊ธฐ ์œ„ํ•ด CIO๊ฐ€ ๊ฐ•๋ ฅํ•œ AI ๊ฑฐ๋ฒ„๋„Œ์Šค์™€ ๋น„์šฉ ํ†ต์ œ ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ๊ตฌ์ถ•ํ•ด์•ผ ํ•œ๋‹ค๋Š” ์š”๊ตฌ๊ฐ€ ์ ์ฐจ ์ปค์ง€๊ณ  ์žˆ์Œ์ด ํ™•์ธ๋๋‹ค.

๋‹จ์ˆœ ํ™•์žฅ์„ ๋„˜์–ด ์ฒด๊ณ„์  ํ™•์žฅ์œผ๋กœ

๊ธฐ์ˆ  ์ „๋žต์€ ๋‹จ์ˆœํžˆ ๊ทœ๋ชจ๋ฅผ ๋Š˜๋ฆฌ๋Š” ๋ฐฉ์‹์ด ์•„๋‹ˆ๋ผ, ๊ฒฝ์ œ์  ๊ธฐ์ค€์— ๋”ฐ๋ผ ๋ชจ๋ธ์„ ์ง€์†์ ์œผ๋กœ ๋ชจ๋‹ˆํ„ฐ๋งํ•˜๊ณ  ๋ถ„์„ํ•˜๋ฉฐ ์ตœ์ ํ™”ํ•˜๊ณ  ๋ฐฐํฌํ•˜๋Š” โ€˜์Šค๋งˆํŠธ ํ™•์žฅ(scale-smart)โ€™ ์ฒ ํ•™์„ ๊ตฌํ˜„ํ•ด์•ผ ํ•œ๋‹ค. ์ด๋Š” ๋ชจ๋ธ์ด ์š”๊ตฌํ•˜๋Š” ์„ฑ๋Šฅ๊ณผ ์ธํ”„๋ผ๊ฐ€ ์ œ๊ณตํ•  ์ˆ˜ ์žˆ๋Š” ์—ญ๋Ÿ‰์„ ์ •๋ฐ€ํ•˜๊ฒŒ ๋งž์ถ”๋Š” ์ง€๋Šฅํ˜• ์šด์˜ ์ฒด๊ณ„๋ฅผ ์˜๋ฏธํ•œ๋‹ค. ์ด๋Ÿฌํ•œ ์šด์˜ ๋ฐฉ์‹์œผ๋กœ์˜ ์ „ํ™˜์ด ์ค‘์š”ํ•œ ์ด์œ ๋Š”, ์ตœ๊ทผ AI ํ˜์‹ ์˜ ํ•ต์‹ฌ์œผ๋กœ ๊ผฝํžˆ๋Š” 2๊ฐ€์ง€ ๊ธฐ์ˆ ์„ ๋„์ž…ํ•˜๋ ค๋ฉด ์ž์›์„ ํšจ์œจ์ ์œผ๋กœ ๋ฐฐ๋ถ„ํ•˜๊ณ  ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋Š” ์ฒด๊ณ„๊ฐ€ ํ•„์ˆ˜์ด๊ธฐ ๋•Œ๋ฌธ์ด๋‹ค.

  • ์†Œ๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ(SLM): ๊ธฐ์—… ์ „์šฉ ๋ฐ์ดํ„ฐ๋กœ ์ •๊ตํ•˜๊ฒŒ ์กฐ์ •๋œ SLM์€ ๋ฒ”์šฉ ๋Œ€๊ทœ๋ชจ ๋ชจ๋ธ๋ณด๋‹ค ํŠน์ • ๊ธฐ์—… ์—…๋ฌด์—์„œ ํ›จ์”ฌ ๋†’์€ ์ •ํ™•๋„์™€ ๋ฌธ๋งฅ ์ ํ•ฉ์„ฑ์„ ์ œ๊ณตํ•œ๋‹ค. ๋ชจ๋ธ์ด ์ž‘๊ธฐ ๋•Œ๋ฌธ์— ๋น„์šฉ์ด ์ ˆ๊ฐ๋  ๋ฟ ์•„๋‹ˆ๋ผ, ๋†’์€ ์ •๋ฐ€๋„๋กœ ์˜ค๋ฅ˜๋ฅผ ์ค„์—ฌ ์ถ”๊ฐ€ ๋น„์šฉ์„ ๋ฐฉ์ง€ํ•  ์ˆ˜ ์žˆ๋‹ค. ์ผ๋ถ€ ์—ฐ๊ตฌ์—์„œ๋„ SLM์„ ๋„์ž…ํ•œ ๊ธฐ์—…์ด ๋ฒ”์šฉ ๋ชจ๋ธ์„ ์‚ฌ์šฉํ•œ ๊ธฐ์—…๋ณด๋‹ค ๋” ๋‚˜์€ ์ •ํ™•๋„์™€ ๋น ๋ฅธ ROI๋ฅผ ์ฐฝ์ถœํ•œ ๊ฒƒ์œผ๋กœ ๋‚˜ํƒ€๋‚ฌ๋‹ค. ๊ฐ€ํŠธ๋„ˆ๋Š” 2027๋…„๊นŒ์ง€ ๊ธฐ์—…์ด ์—…๋ฌด ํŠนํ™” SLM์„ ๋ฒ”์šฉ LLM๋ณด๋‹ค 3๋ฐฐ ๋” ๋งŽ์ด ํ™œ์šฉํ•  ๊ฒƒ์œผ๋กœ ๋‚ด๋‹ค๋ดค๋‹ค.
  • ์—์ด์ „ํ‹ฑ ์›Œํฌํ”Œ๋กœ์šฐ: ์ฐจ์„ธ๋Œ€ ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์€ ๋‹จ์ผ ์‚ฌ์šฉ์ž ์š”์ฒญ์ด ์—ฌ๋Ÿฌ ๋ชจ๋ธ๋กœ ์—ฐ์‡„์ ์œผ๋กœ ์ „๋‹ฌ๋˜๋Š” ์—์ด์ „ํ‹ฑ AI๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ํ•œ๋‹ค. ๋‹ค์ค‘ ๋ชจ๋ธ์— ๊ธฐ๋ฐ˜ํ•œ ํ”„๋กœ์„ธ์Šค๋ฅผ ์šด์˜ํ•˜๋ ค๋ฉด, ํ‚ค ๊ฐ’(KV) ์บ์‹œ ์œ„์น˜ ๊ธฐ๋ฐ˜ ๋ผ์šฐํŒ…, ์ž๋™ ํ”„๋ฆฌํ•„/๋””์ฝ”๋”ฉ ๋ถ„๋ฆฌ, ํ”Œ๋ž˜์‹œ ์–ดํ…์…˜, ์–‘์žํ™”, ์ถ”์ธก ๋””์ฝ”๋”ฉ, ์ด๊ธฐ์ข… GPU ๋ฐ CPU ๊ฐ„ ๋ชจ๋ธ ์ƒค๋”ฉ ๋“ฑ ๋ณต์žกํ•œ ์ตœ์ ํ™”๋ฅผ ์ž๋™์œผ๋กœ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ๋Š” ์ง€๋Šฅํ˜• ํ”Œ๋žซํผ์ด ํ•„์š”ํ•˜๋‹ค. ์š”์•ฝํ•˜๋ฉด, ์ด๋Ÿฐ ๊ธฐ์ˆ ์€ ๋ณต์žกํ•œ AI ์ž‘์—…์˜ ์ง€์—ฐ๊ณผ ๋น„์šฉ์„ ํฌ๊ฒŒ ์ค„์—ฌ์ฃผ๋Š” ํ•ต์‹ฌ ์š”์†Œ๋‹ค.

SLM์ด๋“  ์—์ด์ „ํ‹ฑ ์›Œํฌํ”Œ๋กœ๋“ , ๋˜๋Š” ์–ด๋А ๋ชจ๋ธ์ด๋“  ์ถ”๋ก ์ด ์‹คํ–‰๋˜๋Š” ์ˆœ๊ฐ„ ๋น„์šฉ ํšจ์œจ์„ฑ์„ ํ™•๋ณดํ•˜๋ ค๋ฉด ๋ชจ๋“  ์š”์ฒญ์ด ๋น„์šฉ ์ •์ฑ…์— ๋”ฐ๋ผ ์ž๋™์œผ๋กœ ์ตœ์  ๊ฒฝ๋กœ๋กœ ๋ผ์šฐํŒ…๋˜๊ณ , ํ•˜๋“œ์›จ์–ด ํŠน์„ฑ์— ๋งž์ถฐ ์ง€์†์ ์œผ๋กœ ์‹คํ–‰ ๋ฐฉ์‹์ด ์กฐ์ •๋ผ์•ผ ํ•œ๋‹ค. ์ด๋Ÿฐ ๊ตฌ์กฐ๋กœ ์ตœ์ ํ™”๋  ๋•Œ๋งŒ ๋ฐฑ๋งŒ ํ† ํฐ๋‹น ๋น„์šฉ์„ ๋‘ ์ž๋ฆฟ์ˆ˜ ์ˆ˜์ค€์œผ๋กœ ๋‚ฎ์ถœ ์ˆ˜ ์žˆ๋‹ค. ์ด๋ฅผ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๋Š” ์œ ์ผํ•œ ๊ธฐ๋ฐ˜์€ ๊ธฐ์—… ์ „๋ฐ˜์˜ ์ถ”๋ก ์„ ์ผ๊ด€๋˜๊ฒŒ ์šด์˜ํ•  ์ˆ˜ ์žˆ๋Š” ์ค‘์•™ ํ†ตํ•ฉ ํ”Œ๋žซํผ์ด๋‹ค.

๊ธฐ์กด AI ์ถ”๋ก ์˜ ๋น„ํšจ์œจ ํ•ด๊ฒฐ

๊ธฐ์กด ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ์ธํ”„๋ผ๋ฅผ ์šด์˜ํ•˜๋˜ ๋ฐฉ์‹, ์ฆ‰ ํ•„์ž๊ฐ€ โ€˜๋‹จ์ˆœ ํ™•์žฅโ€™์ด๋ผ๊ณ  ๋ถ€๋ฅด๋Š” ์ ‘๊ทผ์€ ์ง€์†์ ์ธ AI ์ถ”๋ก  ํ™˜๊ฒฝ์—์„œ๋Š” ์ œ๋Œ€๋กœ ์ž‘๋™ํ•˜์ง€ ์•Š๋Š”๋‹ค. ์ด ๋ฐฉ์‹์€ ์˜ค๋Š˜๋‚  CIO๊ฐ€ ํ•„์š”๋กœ ํ•˜๋Š” ์ถ”๋ก  ํ”Œ๋žซํผ ๊ตฌ์ถ•์—๋„ ํ™œ์šฉ๋˜๊ธฐ ์–ด๋ ต๋‹ค. ๊ทธ๋™์•ˆ ๊ธฐ์—…์€ ์ „์šฉ ๋Œ€๊ทœ๋ชจ ํด๋Ÿฌ์Šคํ„ฐ๋ฅผ ๋ฏธ๋ฆฌ ๊ณผ๋‹คํ•˜๊ฒŒ ํ™•๋ณดํ•˜๊ณ , ์ตœ์‹  GPU๋ฅผ ๋„์ž…ํ•˜๋ฉฐ, ๊ณ ๋น„์šฉ ํ•™์Šต ํ™˜๊ฒฝ์„ ์ถ”๋ก  ๋‹จ๊ณ„์—์„œ๋„ ๊ทธ๋Œ€๋กœ ํ™œ์šฉํ•ด ์™”๋‹ค.

๊ทธ๋Ÿฌ๋‚˜ ์ด๋Š” ์ตœ์†Œ 2๊ฐ€์ง€ ์ด์œ ์—์„œ ๊ทผ๋ณธ์ ์œผ๋กœ ๋น„ํšจ์œจ์ ์ด๋‹ค.

  • ์ถ”๋ก  ์ž‘์—…์€ ๋ถ€ํ•˜ ๋ณ€๋™์ด ๋งค์šฐ ํฌ๊ณ  ์œ ํœด ์‹œ๊ฐ„์ด ๊ธธ๋‹ค. ์žฅ์‹œ๊ฐ„ ์‹คํ–‰๋˜๋Š” ํ›ˆ๋ จ๊ณผ ๋‹ฌ๋ฆฌ, ์ถ”๋ก  ์š”์ฒญ์€ ๊ฐ‘์ž‘์Šค๋Ÿฝ๊ฒŒ ๋ชฐ๋ฆฌ๊ธฐ๋„ ํ•˜๊ณ  ์˜ˆ์ธกํ•˜๊ธฐ ์–ด๋ ค์šฐ๋ฉฐ, ๊ทธ ์‚ฌ์ด์— ๊ณต๋ฐฑ์ด ๊ธธ๊ฒŒ ๋ฐœ์ƒํ•œ๋‹ค. ์ด๋Ÿฐ ๊ฐ„ํ—์  ์š”์ฒญ์„ ์ฒ˜๋ฆฌํ•˜๊ธฐ ์œ„ํ•ด ๋Œ€๊ทœ๋ชจ ํด๋Ÿฌ์Šคํ„ฐ๋ฅผ ์œ ์ง€ํ•œ๋‹ค๋ฉด ๊ฒฐ๊ตญ ๋ง‰๋Œ€ํ•œ ์šฉ๋Ÿ‰์„ ๋†€๋ฆฌ๋Š” ์…ˆ์ด ๋˜๊ณ , ํ™œ์šฉ๋ฅ ์€ ๋–จ์–ด์ง€๋ฉฐ ์žฌ๋ฌด ๋ถ€์„œ๋Š” ์ด์— ๋Œ€ํ•ด ์˜๋ฌธ์„ ์ œ๊ธฐํ•˜๊ฒŒ ๋œ๋‹ค. ์ง€๊ธˆ ์ค‘์š”ํ•œ ๋น„์šฉ ์ง€ํ‘œ๋Š” ์ด๋ก ์  ์ฒ˜๋ฆฌ๋Ÿ‰์ด ์•„๋‹ˆ๋ผ ๋ฐฑ๋งŒ ํ† ํฐ๋‹น ์‹ค์ œ ๋น„์šฉ์ด๋‹ค. ๊ฐ€ํŠธ๋„ˆ ์กฐ์‚ฌ์—์„œ๋„ ์ƒ์„ฑํ˜• AI์˜ ์˜ˆ์ธกํ•˜๊ธฐ ์–ด๋ ค์šด ๋น„์šฉ ๊ธ‰์ฆ์„ ๊ด€๋ฆฌํ•˜๋Š” ๊ฒƒ์ด CIO์˜ ์ตœ์šฐ์„  ๊ณผ์ œ ์ค‘ ํ•˜๋‚˜๋กœ ๊ผฝํ˜”๋‹ค. ์ตœ์ ํ™”์˜ ๋ชฉํ‘œ๋Š” ์ด๋ก ์  ์„ฑ๋Šฅ์ด ์•„๋‹ˆ๋ผ ๊ฒฝ์ œ์„ฑ์ด๋‹ค.
  • ๋˜ํ•œ AI ์ถ”๋ก ์€ ํƒœ์ƒ์ ์œผ๋กœ ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ๋ฐฐํฌ ํ™˜๊ฒฝ์„ ์ „์ œ๋กœ ํ•œ๋‹ค. ์ถ”๋ก ์„ ์ค‘์•™ ์ง‘์ค‘์‹์˜ ๋‹จ์ผํ•˜๊ณ  ๊ท ์งˆํ•œ ํ™˜๊ฒฝ์—์„œ๋งŒ ์ฒ˜๋ฆฌํ•˜๊ธฐ๋Š” ์‚ฌ์‹ค์ƒ ๋ถˆ๊ฐ€๋Šฅํ•˜๋‹ค. ๊ธˆ์œต, ์˜๋ฃŒ์ฒ˜๋Ÿผ ๊ทœ์ œ๊ฐ€ ๊ฐ•ํ•œ ์‚ฐ์—…์ด๋‚˜, ๋‚ด๋ถ€ ๊ณ ์œ  ๋ฐ์ดํ„ฐ๋ฅผ ๊ธฐ๋ฐ˜์œผ๋กœ ์šด์˜๋˜๋Š” ์—…๋ฌด์—์„œ๋Š” ๋ฐ์ดํ„ฐ๊ฐ€ ๋ณด์•ˆ ๊ฒฝ๊ณ„๋ฅผ ๋ฒ—์–ด๋‚˜์„  ์•ˆ ๋œ๋‹ค. ๋”ฐ๋ผ์„œ ์ถ”๋ก ์€ ์˜จํ”„๋ ˆ๋ฏธ์Šค, ์—ฃ์ง€, ๋˜๋Š” ๋ณด์•ˆ์ด ๋ณด์žฅ๋œ ์ฝœ๋กœ์ผ€์ด์…˜ ๋“ฑ ๋‹ค์–‘ํ•œ ํ™˜๊ฒฝ์—์„œ ์ˆ˜ํ–‰๋ผ์•ผ ํ•œ๋‹ค. ์ค‘์š”ํ•œ ์›Œํฌ๋กœ๋“œ๋ฅผ ์ผ๋ฐ˜์ ์ธ ํด๋ผ์šฐ๋“œ API ์—”๋“œํฌ์ธํŠธ๋กœ ๋ฐ€์–ด๋„ฃ๋Š” ๋ฐฉ์‹์€ ์ด๋Ÿฌํ•œ ๊ทœ์ œ ๋ฐ ๋ณด์•ˆ ์š”๊ฑด์„ ์ถฉ์กฑํ•˜๊ธฐ ์–ด๋ ค์›Œ, ๊ธฐ์—…์€ ์ž์—ฐ์Šค๋Ÿฝ๊ฒŒ ํ•˜์ด๋ธŒ๋ฆฌ๋“œ ๋ฐ ์—ฃ์ง€ ์•„ํ‚คํ…์ฒ˜๋กœ ์ด๋™ํ•˜๊ณ  ์žˆ๋‹ค. ์—ฌ๊ธฐ์— ๋”ํ•ด ํ•˜๋“œ์›จ์–ด ์—ญ์‹œ CPU, GPU, DPU ๋“ฑ ๋‹ค์–‘ํ•œ ํ”„๋กœ์„ธ์„œ๊ฐ€ ํ˜ผ์žฌ๋œ ์ด๊ธฐ์ข… ํ™˜๊ฒฝ์ด๊ธฐ ๋•Œ๋ฌธ์— ํ”Œ๋žซํผ์€ ์ด๋ฅผ ๋ชจ๋‘ ๋งค๋„๋Ÿฝ๊ฒŒ ๊ด€๋ฆฌํ•  ์ˆ˜ ์žˆ์–ด์•ผ ํ•œ๋‹ค.

CIO์˜ ํ•„์ˆ˜ ๊ณผ์ œ๋Š” โ€˜AI ์ถ”๋ก  ํ”Œ๋žซํผ์˜ ์™„์„ฑโ€™

ํ†ตํ•ฉ ํ”Œ๋žซํผ์˜ ๋ชฉ์ ์€ ํ•˜๋‚˜์˜ ๋ชจ๋ธ๋กœ ๊ฐ•์ œ ํ†ต์ผํ•˜๋Š” ๊ฒƒ์ด ์•„๋‹ˆ๋‹ค. ๊ธฐ์—…์ด ์š”๊ตฌํ•˜๋Š” ๋ณด์•ˆ๊ณผ ๋น„์šฉ ๊ด€๋ฆฌ ๊ธฐ์ค€์„ ์ถฉ์กฑํ•˜๋ฉด์„œ ํ›จ์”ฌ ๋‹ค์–‘ํ•œ ๋ชจ๋ธ, ์—์ด์ „ํŠธ, ์• ํ”Œ๋ฆฌ์ผ€์ด์…˜์„ ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ํ•˜๋Š” ๊ฑฐ๋ฒ„๋„Œ์Šค ๊ณ„์ธต์„ ๋งˆ๋ จํ•˜๋Š” ๋ฐ ์žˆ๋‹ค.

โ€˜๋‹จ์ˆœ ํ™•์žฅโ€™์—์„œ โ€˜์Šค๋งˆํŠธ ํ™•์žฅโ€™์œผ๋กœ์˜ ์ „ํ™˜์€ ๊ธฐ์ˆ  ๋ฆฌ๋”์—๊ฒŒ ์ฃผ์–ด์ง„ ํ•ต์‹ฌ ๊ณผ์ œ๋‹ค. AI์˜ ๋ฏธ๋ž˜ ๊ฐ€์น˜๋Š” ํ›ˆ๋ จํ•œ ๋ชจ๋ธ์ด ์•„๋‹ˆ๋ผ, ์ถ”๋ก  ์šด์˜์„ ํ†ตํ•ด ์–ผ๋งˆ๋‚˜ ์•ˆ์ •์ ์ธ ๋งˆ์ง„์„ ํ™•๋ณดํ•˜๋Š”๊ฐ€์— ๋‹ฌ๋ ค์žˆ๋‹ค.

๋ชจ๋“  ๊ธฐ์ˆ  ๋ฆฌ๋”๋Š” AI ์†์ต ์„ผํ„ฐ์˜ ํ”Œ๋žซํผ ์†Œ์œ ์ž์ด์ž ์žฌ๋ฌด ์„ค๊ณ„์ž๋กœ์„œ ์—ญํ• ์„ ์ˆ˜ํ–‰ํ•ด์•ผ ํ•œ๋‹ค. ์ด ๊ตฌ์กฐ์  ๋ณ€ํ™”๊ฐ€ ์ด๋ฃจ์–ด์ ธ์•ผ๋งŒ ๋ฐ์ดํ„ฐ ์‚ฌ์ด์–ธ์ŠคํŒ€์ด ๋ณด์•ˆยท๊ทœ์ œ ์ค€์ˆ˜ยท๋น„์šฉ์ด ์ตœ์ ํ™”๋œ ๊ธฐ๋ฐ˜ ์œ„์—์„œ ๊ธฐ์กด ์†๋„๋ฅผ ์œ ์ง€ํ•˜๋ฉฐ ํ˜์‹ ์„ ์ด์–ด๊ฐˆ ์ˆ˜ ์žˆ๋‹ค.

ํ”Œ๋žซํผ์„ ์ •์ฐฉ์‹œํ‚ค๊ณ  ์Šค๋งˆํŠธ ํ™•์žฅ ์ „๋žต์„ ๋„์ž…ํ•˜๋ฉด, AI ๋น„์šฉ์ด ๊ฑท์žก์„ ์ˆ˜ ์—†์ด ์ฆ๊ฐ€ํ•˜๋Š” ์ƒํ™ฉ์—์„œ ๋ฒ—์–ด๋‚˜ ์ง€์† ๊ฐ€๋Šฅํ•œ ์ˆ˜์ต ๊ธฐ๋ฐ˜ ๊ฒฝ์Ÿ๋ ฅ์„ ํ™•๋ณดํ•  ์ˆ˜ ์žˆ๋‹ค. ๋‚จ์€ ์„ ํƒ์ง€๋Š” ๋ถ„๋ช…ํ•˜๋‹ค. ๋ถ„์‚ฐ๋œ AI ํ™˜๊ฒฝ์˜ ๋น„์šฉ๊ณผ ํ˜ผ๋ž€์„ ๊ณ„์†ํ•ด์„œ ์ˆ˜๋™์ ์œผ๋กœ ๊ด€๋ฆฌํ•  ๊ฒƒ์ธ์ง€, ์•„๋‹ˆ๋ฉด ์ถ”๋ก ์„ ์ˆ˜์ต ์ฐฝ์ถœ ๊ธฐ๋ฐ˜์œผ๋กœ ์ „ํ™˜ํ•˜๋Š” AI ์†์ต ์„ผํ„ฐ๋ฅผ ๊ตฌ์ถ•ํ•ด ์žฅ๊ธฐ์  ๊ฒฝ์Ÿ์šฐ์œ„๋ฅผ ํ™•๋ณดํ•  ๊ฒƒ์ธ๊ฐ€๋‹ค.
dl-ciokorea@foundryco.com

โŒ
โŒ