โŒ

Normal view

There are new articles available, click to refresh the page.
Before yesterdayMain stream

โ€œ์˜ฌํ•ด ๋ณด์•ˆ, ์ด๊ฒƒ๋งŒ์€ ํ•„์ˆ˜โ€ ๊ธ€๋กœ๋ฒŒ ๋ฆฌ๋”๊ฐ€ ๊ผฝ์€ 2026๋…„ ๋ณด์•ˆ ์šฐ์„  ์ˆœ์œ„

20 January 2026 at 00:47

2026๋…„์„ ์•ž๋‘๊ณ  CISO์™€ ๋Š์ž„์—†์ด ์ง„ํ™”ํ•˜๋Š” ์‚ฌ์ด๋ฒ„ ๊ณต๊ฒฉ์ž ๊ฐ„์˜ ๋Œ€๊ฒฐ์ด ๋‹ค์‹œ ํ•œ ๋ฒˆ ๊ฒฉํ™”๋˜๋Š” ๊ฐ€์šด๋ฐ, ๊ณต๊ฒฉ์ž๋ณด๋‹ค ํ•œ๋ฐœ ์•ž์„œ ์ฃผ๋„๊ถŒ์„ ์œ ์ง€ํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ์น˜๋ฐ€ํ•˜๊ฒŒ ๊ธฐํš๋œ ๊ฐ•๋ ฅํ•œ ์‚ฌ์ด๋ฒ„ ๋ณด์•ˆ ํ”„๋กœ์ ํŠธ๊ฐ€ ํšจ๊ณผ์ ์ธ ๋Œ€์‘์ฑ…์ผ ์ˆ˜ ์žˆ๋‹ค.

๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค๋ถ€ํ„ฐ ์ œ๋กœ ํŠธ๋Ÿฌ์ŠคํŠธ๊นŒ์ง€, ํ–ฅํ›„ 1๋…„ ๋™์•ˆ ๋ชจ๋“  CISO๊ฐ€ ๋„์ž…์„ ๊ฒ€ํ† ํ•ด๋ณผ ๋งŒํ•œ ํ•ต์‹ฌ ์‚ฌ์ด๋ฒ„ ๋ณด์•ˆ ํ”„๋กœ์ ํŠธ 7๊ฐ€์ง€๋ฅผ ์ •๋ฆฌํ–ˆ๋‹ค.

1. AI ์‹œ๋Œ€๋ฅผ ์œ„ํ•œ ์•„์ด๋ดํ‹ฐํ‹ฐ ๋ฐ ์ ‘๊ทผ ๊ด€๋ฆฌ ์ „ํ™˜

AI์™€ ์ž๋™ํ™” ๊ธฐ์ˆ ์ด ์ง„ํ™”ํ•˜๋ฉด์„œ ์ง์›์˜ ์ ‘๊ทผ ๊ถŒํ•œ๋ฟ ์•„๋‹ˆ๋ผ AI ์—์ด์ „ํŠธ์™€ ๋จธ์‹  ํ”„๋กœ์„ธ์Šค์˜ ์•„์ด๋ดํ‹ฐํ‹ฐ๊นŒ์ง€ ๊ด€๋ฆฌํ•˜๋Š” ๊ฒƒ์ด ์ด์ œ ํ•„์ˆ˜์ ์ธ ์‚ฌ์ด๋ฒ„ ๋ณด์•ˆ ์š”์†Œ๋กœ ์ž๋ฆฌ ์žก๊ณ  ์žˆ๋‹ค. ๋”œ๋กœ์ดํŠธ ๋ฏธ๊ตญ ์‚ฌ์ด๋ฒ„ ์•„์ด๋ดํ‹ฐํ‹ฐ ๋ถ€๋ฌธ ๋ฆฌ๋”์ธ ์•ค์„œ๋‹ˆ ๋ฒ„๊ทธ๋Š” ์ด๋Ÿฌํ•œ ๋ณ€ํ™”๋ฅผ ๋ณด์•ˆ์˜ ํ•ต์‹ฌ ๊ณผ์ œ๋กœ ์งš์—ˆ๋‹ค.

๋ฒ„๊ทธ๋Š” โ€œํŠนํžˆ ์—์ด์ „ํ‹ฑ AI๋ฅผ ์ค‘์‹ฌ์œผ๋กœ ํ•œ AI์˜ ๋น ๋ฅธ ๋ฐœ์ „์ด ๋งŽ์€ ๋ณด์•ˆ ๋ฆฌ๋”๋กœ ํ•˜์—ฌ๊ธˆ ์•„์ด๋ดํ‹ฐํ‹ฐ ๊ด€๋ฆฌ ์ „๋žต์„ ๋‹ค์‹œ ์ƒ๊ฐํ•˜๊ฒŒ ๋งŒ๋“ค๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ์‚ฌ๋žŒ๊ณผ ๋น„์ธ๊ฐ„ ์•„์ด๋ดํ‹ฐํ‹ฐ๋ฅผ ๋ชจ๋‘ ์•„์šฐ๋ฅด๋Š” ๋ณด๋‹ค ์ •๊ตํ•œ ์•„์ด๋ดํ‹ฐํ‹ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค์— ๋Œ€ํ•œ ์š”๊ตฌ๊ฐ€ CISO์™€ CIO๋กœ ํ•˜์—ฌ๊ธˆ ์ฐจ์„ธ๋Œ€ ๋””์ง€ํ„ธ ์ „ํ™˜์„ ๋Œ€๋น„ํ•œ ๋ณด์•ˆ ํ”„๋ ˆ์ž„์›Œํฌ๋ฅผ ์žฌ๊ตฌ์„ฑํ•˜๋„๋ก ์ด๋Œ๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๊ทธ๋Š” ์ƒ์„ฑํ˜• AI์™€ ์—์ด์ „ํ‹ฑ AI๊ฐ€ ์ƒˆ๋กœ์šด ๋น„์ฆˆ๋‹ˆ์Šค ๋ชจ๋ธ๊ณผ ๋” ๋†’์€ ์ˆ˜์ค€์˜ ์ž์œจ์„ฑ์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๋Š” ๋งŒํผ, ์กฐ์ง์ด ์•„์ด๋ดํ‹ฐํ‹ฐ ๋ฐ ์ ‘๊ทผ ๊ด€๋ฆฌ(IAM) ํ”„๋กœ๊ทธ๋žจ์„ ์„ ์ œ์ ์œผ๋กœ ํ˜„๋Œ€ํ™”ํ•˜๋Š” ๊ฒƒ์ด ์ค‘์š”ํ•˜๋‹ค๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค. ๋ชจ๋“  ๋””์ง€ํ„ธ ์•„์ด๋ดํ‹ฐํ‹ฐ ์ „๋ฐ˜์˜ ์ ‘๊ทผ์„ ๋ณดํ˜ธํ•˜๋Š” ๊ฒƒ์€ ๋ฏผ๊ฐํ•œ ๋ฐ์ดํ„ฐ ๋ณดํ˜ธ์™€ ๊ทœ์ œ ์ค€์ˆ˜, ์šด์˜ ํšจ์œจ์„ฑ ํ™•๋ณด๋ฅผ ์œ„ํ•ด ํ•„์ˆ˜์ ์ด๋ผ๋Š” ์„ค๋ช…์ด๋‹ค.

๋ฒ„๊ทธ๋Š” ๋ผ์ดํ”„์‚ฌ์ดํด ๊ด€๋ฆฌ, ๊ฐ•๋ ฅํ•œ ์ธ์ฆ, ์ •๋ฐ€ํ•œ ์—ญํ•  ๋ฐ ์ •์ฑ… ๊ธฐ๋ฐ˜ ์ ‘๊ทผ ์ œ์–ด์™€ ๊ฐ™์€ IAM ์—ญ๋Ÿ‰์„ ๊ณ ๋„ํ™”ํ•˜๋ฉด ๋น„์ธ๊ฐ€ ์ ‘๊ทผ์„ ์ฐจ๋‹จํ•˜๊ณ  ํƒˆ์ทจ๋œ ์ž๊ฒฉ ์ฆ๋ช…์œผ๋กœ ์ธํ•œ ์œ„ํ—˜์„ ์ค„์ผ ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค.

๋˜ํ•œ ์ด๋Ÿฌํ•œ ํ†ต์ œ๋ฅผ ๋น„์ธ๊ฐ„ ์•„์ด๋ดํ‹ฐํ‹ฐ๊นŒ์ง€ ํ™•์žฅํ•˜๋ฉด ์‹œ์Šคํ…œ์ด๋‚˜ ๋ฐ์ดํ„ฐ์™€ ์ƒํ˜ธ์ž‘์šฉํ•˜๋Š” ๋ชจ๋“  ์ฃผ์ฒด๋ฅผ ์ ์ ˆํžˆ ๊ด€๋ฆฌํ•  ์ˆ˜ ์žˆ์œผ๋ฉฐ, ์ •๊ธฐ์ ์ธ ์ ‘๊ทผ ๊ถŒํ•œ ๊ฒ€ํ† ์™€ ์ง€์†์ ์ธ ๊ต์œก์„ ๋ณ‘ํ–‰ํ•  ๊ฒฝ์šฐ ์ •๋ณด ๋ณดํ˜ธ ์ˆ˜์ค€์„ ๋†’์ด๊ณ  ๊ณ ๋„ํ™”๋œ AI ๊ธฐ์ˆ ์„ ๋ณด๋‹ค ์•ˆ์ „ํ•˜๊ฒŒ ๋„์ž…ํ•˜๋Š” ๋ฐ ๋„์›€์ด ๋œ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

2. ์ด๋ฉ”์ผ ๋ณด์•ˆ ๊ฐ•ํ™”

์นด๋„ค๊ธฐ๋ฉœ๋ก ๋Œ€ํ•™๊ต CISO์ธ ๋ฉ”๋ฆฌ ์•ค ๋ธ”๋ ˆ์–ด๋Š” ํ”ผ์‹ฑ์ด ์—ฌ์ „ํžˆ ์ž๊ฒฉ ์ฆ๋ช…์„ ํƒˆ์ทจํ•˜๊ณ  ํ”ผํ•ด์ž๋ฅผ ์†์ด๋Š” ์ฃผ์š” ๊ณต๊ฒฉ ๊ฒฝ๋กœ๋กœ ํ™œ์šฉ๋˜๊ณ  ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” ์œ„ํ˜‘ ํ–‰์œ„์ž๋“ค์ด ๋ฉ”์ผ ์„œ๋น„์Šค ์ œ๊ณต์—…์ฒด์˜ ํƒ์ง€ ๊ธฐ๋Šฅ์„ ํšจ๊ณผ์ ์œผ๋กœ ํšŒํ”ผํ•  ์ˆ˜ ์žˆ์„ ๋งŒํผ ์ ์  ๋” ์ •๊ตํ•œ ํ”ผ์‹ฑ ๊ณต๊ฒฉ์„ ๋งŒ๋“ค์–ด๋‚ด๊ณ  ์žˆ๋‹ค๊ณ  ๊ฒฝ๊ณ ํ–ˆ๋‹ค.

๋ธ”๋ ˆ์–ด๋Š” โ€œ๊ธฐ์กด์˜ ๋‹ค์ค‘์š”์†Œ ์ธ์ฆ ๊ธฐ๋ฒ•์€ ์ด์ œ ๋ฐ˜๋ณต์ ์œผ๋กœ ๋ฌด๋ ฅํ™”๋˜๊ณ  ์žˆ์œผ๋ฉฐ, ๊ณต๊ฒฉ์ž๋“ค์€ ์นจํˆฌ์— ์„ฑ๊ณตํ•œ ์ดํ›„ ์ด๋ฅผ ๋น ๋ฅด๊ฒŒ ์ˆ˜์ตํ™”ํ•˜๋Š” ๋‹จ๊ณ„๋กœ ์ด๋™ํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.

์ด์ฒ˜๋Ÿผ ๊ฐˆ์ˆ˜๋ก ๋Œ€์‘์ด ์–ด๋ ค์›Œ์ง€๋Š” ์ด๋ฉ”์ผ ๋ณด์•ˆ ํ™˜๊ฒฝ ์†์—์„œ ๋ธ”๋ ˆ์–ด๋Š” CISO๊ฐ€ ๋ณด์•ˆ ํ”„๋กœ์ ํŠธ๋ฅผ ์ถ”์ง„ํ•˜๋Š” ๊ณผ์ •์—์„œ ์™ธ๋ถ€ ์ „๋ฌธ ์กฐ์ง์˜ ์ง€์›์„ ๊ฒ€ํ† ํ•  ํ•„์š”๊ฐ€ ์žˆ๋‹ค๊ณ  ์กฐ์–ธํ–ˆ๋‹ค. ์‹ค์ œ๋กœ ๊ทธ๊ฐ€ ์ ‘์ด‰ํ•œ ์—ฌ๋Ÿฌ ๋ฒค๋”๋Š” ์ œ์•ˆ์š”์ฒญ์„œ(RFP)์— ์‘๋‹ตํ•˜๋Š” ํ•œํŽธ, ์ตœ์‹  ๋ณด์•ˆ ์—ญ๋Ÿ‰์„ ์‹œํ—˜ ์ ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ํ…Œ์ŠคํŠธ ํ™˜๊ฒฝ์„ ์ œ๊ณตํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ์ „ํ–ˆ๋‹ค.

3. AI๋ฅผ ํ™œ์šฉํ•œ ์ฝ”๋“œ ์ทจ์•ฝ์  ํƒ์ง€

์‹œ์Šค์ฝ” AI ์—ฐ๊ตฌ์› ์•„๋งŒ ํ”„๋ฆฌ์–€์Šˆ๋Š” ์ž์›์ด ์ œํ•œ๋œ ํ™˜๊ฒฝ์—์„œ๋„ ํšจ๊ณผ์ ์œผ๋กœ ๋™์ž‘ํ•  ์ˆ˜ ์žˆ๋Š” ์†Œํ˜• ์–ธ์–ด ๋ชจ๋ธ(SLM)์„ ํ™œ์šฉํ•ด ์ž์œจ์ ์œผ๋กœ ์ทจ์•ฝ์ ์„ ํƒ์ƒ‰ํ•˜๋Š” ์—์ด์ „ํŠธ๋ฅผ ๊ฐœ๋ฐœํ•˜๊ณ  ์žˆ๋‹ค.

ํ”„๋ฆฌ์–€์Šˆ๋Š” ์‚ฌ์ด๋ฒ„ ๋ณด์•ˆ์ด ๋ณธ์งˆ์ ์œผ๋กœ ๊ธด ๋งฅ๋ฝ์„ ๋‹ค๋ค„์•ผ ํ•˜๋Š” ์˜์—ญ์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ตœ์‹  ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ(LLM)์ด ์ด๋ฅผ ์ฒ˜๋ฆฌํ•  ์ˆ˜๋Š” ์žˆ์ง€๋งŒ, ๋น„์šฉ์ด๋‚˜ ์ง€์—ฐ ์‹œ๊ฐ„ ์ธก๋ฉด์—์„œ ์ƒ๋‹นํ•œ ๋ถ€๋‹ด์ด ๋”ฐ๋ฅธ๋‹ค๋Š” ๊ฒƒ์ด๋‹ค. ๊ทธ๋Š” โ€œ์กฐ์ง์˜ ์ฝ”๋“œ๋ฒ ์ด์Šค๋Š” ๋ณดํ†ต ์ˆ˜์ฒœ ๊ฐœ ํŒŒ์ผ๊ณผ ์ˆ˜๋ฐฑ๋งŒ ์ค„์˜ ์ฝ”๋“œ๋กœ ๊ตฌ์„ฑ๋œ๋‹คโ€๋ผ๋ฉฐ โ€œํŠน์ • ์ทจ์•ฝ์ ์„ ์ฐพ์•„์•ผ ํ•  ๋•Œ ๋ชจ๋“  ์ฝ”๋“œ๋ฅผ ๋Œ€ํ˜• ๋ชจ๋ธ์— ์ž…๋ ฅํ•˜๋ฉด ๊ฐ๋‹นํ•˜๊ธฐ ์–ด๋ ค์šธ ์ •๋„๋กœ ๋น„์šฉ์ด ์ปค์ง€๊ฑฐ๋‚˜, ์•„์˜ˆ ๋งฅ๋ฝ ํ•œ๊ณ„๋ฅผ ์ดˆ๊ณผํ•˜๋Š” ๋ฌธ์ œ๊ฐ€ ๋ฐœ์ƒํ•œ๋‹คโ€๋ผ๊ณ  ๋ฐํ˜”๋‹ค.

ํ”„๋ฆฌ์–€์Šˆ๋Š” ์ด ํ”„๋กœ์ ํŠธ๊ฐ€ ๋Œ€๋ถ€๋ถ„์˜ ๋ณด์•ˆ ๋ถ„์„๊ฐ€๊ฐ€ ์ทจ์•ฝ์ ์„ ์ฐพ๋Š” ๋ฐฉ์‹๊ณผ ์œ ์‚ฌํ•œ ์ ‘๊ทผ์„ ๊ตฌํ˜„ํ•˜๋Š” ๋ฐ ๋ชฉ์ ์ด ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ž ์žฌ์ ์ธ ์ทจ์•ฝ ์ง€์ ์„ ์ถ”๋ก ํ•œ ๋’ค ํ•ด๋‹น ์˜์—ญ์„ ํƒ์ƒ‰ํ•˜๊ณ , ๊ด€๋ จ ์ฝ”๋“œ๋ฅผ ๊ฐ€์ ธ์™€ ๋ถ„์„ํ•˜๋Š” ๊ณผ์ •์„ ๋ฐ˜๋ณตํ•ด ์•ฝ์ ์„ ์ฐพ์•„๋‚ด๋Š” ๋ฐฉ์‹์ด๋‹ค. ๊ทธ๋Š” โ€œ์—ฐ๊ตฌ๋ฅผ ํ†ตํ•ด ์ด ์ ‘๊ทผ๋ฒ•์ด ํšจ๊ณผ์ ์ด๋ผ๋Š” ์ ์€ ์ด๋ฏธ ํ™•์ธํ–ˆ๋‹คโ€๋ผ๋ฉฐ โ€œ2026๋…„์—๋Š” ์ด๋ฅผ ํ™•์žฅํ•ด ์‹ค์ œ ํ™˜๊ฒฝ์—์„œ์˜ ์ ์šฉ ๊ฐ€๋Šฅ์„ฑ์„ ์‹ค์งˆ์ ์œผ๋กœ ๊ฒ€์ฆํ•˜๊ณ ์ž ํ•œ๋‹คโ€๋ผ๊ณ  ์ „ํ–ˆ๋‹ค.

์ด๋ฏธ ์นจํˆฌ ํ…Œ์Šคํ„ฐ์™€ ๋ณด์•ˆ ์—ฐ๊ตฌ์›๋“ค์€ ์ƒ์„ฑํ˜• AI๋ฅผ ์ทจ์•ฝ์  ํƒ์ƒ‰์— ํ™œ์šฉํ•ด ์™”์œผ๋ฉฐ, AI ๊ธฐ๋ฐ˜ ๋ฒ„๊ทธ ํ—ŒํŒ…์€ ์ทจ์•ฝ์  ๋ฐœ๊ฒฌ ์†๋„๋ฅผ ๋†’์ด๋Š” ๋™์‹œ์— ๊ทธ ์ ‘๊ทผ์„ฑ์„ ํ™•๋Œ€ํ•˜๋Š” ํ๋ฆ„์„ ๋ณด์ด๊ณ  ์žˆ๋‹ค. ์ด๋Š” ํšจ๊ณผ์ ์ธ ๋ฒ„๊ทธ ๋ฐ”์šดํ‹ฐ ํ”„๋กœ๊ทธ๋žจ์„ ์„ค๊ณ„ํ•˜๋Š” ๊ธฐ์ค€์—๋„ ๋ณ€ํ™”๋ฅผ ์ฃผ๊ณ  ์žˆ๋‹ค.

4. ๊ธฐ์—… ์ „๋ฐ˜์˜ AI ๊ฑฐ๋ฒ„๋„Œ์Šค ๋ฐ ๋ฐ์ดํ„ฐ ๋ณดํ˜ธ ๊ฐ•ํ™”

AI ๋ฆฌ์Šคํฌ์™€ ์ž์œจํ˜• ์œ„ํ˜‘์ด ์‚ฌ์ด๋ฒ„ ๋ณด์•ˆ ํ™˜๊ฒฝ์„ ์žฌํŽธํ•˜๋Š” ๊ฐ€์šด๋ฐ, AI ๊ธฐ๋ฐ˜ ์ปค๋ฎค๋‹ˆ์ผ€์ด์…˜ ๋ฐ ํ˜‘์—… ์†”๋ฃจ์…˜ ๊ธฐ์—… ๊ณ ํˆฌ(GoTo)์˜ CISO์ธ ์•„ํ‹ธ๋ผ ํ‡ด๋ขฐํฌ๋Š” ์กฐ์ง ๋‚ด ๋ชจ๋“  AI ๋„๊ตฌ๋ฅผ ์•ˆ์ „ํ•˜๊ฒŒ ๊ด€๋ฆฌยท๋ชจ๋‹ˆํ„ฐ๋งํ•˜๋Š” ํ•œํŽธ, ์Šน์ธ๋˜์ง€ ์•Š์€ ํ”Œ๋žซํผ์„ ์ฐจ๋‹จํ•ด ๋ฐ์ดํ„ฐ ์œ ์ถœ์„ ๋ฐฉ์ง€ํ•˜๋Š” ๋ฐ ์ฃผ๋ ฅํ•˜๊ณ  ์žˆ๋‹ค.

ํ‡ด๋ขฐํฌ๋Š” โ€œ์„ค๊ณ„ ๋‹จ๊ณ„๋ถ€ํ„ฐ ๋ณด์•ˆ์„ ๋‚ด์žฌํ™”ํ•˜๋Š” ์›์น™์„ ์ ์šฉํ•˜๊ณ  ์‚ฌ์ด๋ฒ„ ๋ณด์•ˆ์„ ๋น„์ฆˆ๋‹ˆ์Šค ์ „๋žต๊ณผ ์ •๋ ฌํ•จ์œผ๋กœ์จ ํšŒ๋ณต๋ ฅ๊ณผ ์‹ ๋ขฐ, ๊ทœ์ œ ์ค€์ˆ˜๋ฅผ ๋™์‹œ์— ๊ตฌ์ถ•ํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ์ด๋Ÿฌํ•œ ์š”์†Œ๋Š” AI ์‹œ๋Œ€์˜ ํ•ต์‹ฌ์ ์ธ ์ฐจ๋ณ„ํ™” ์š”์ธโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๋‹ค๋งŒ ๊ทธ๋Š” ์—ฌ๋А ๋Œ€๊ทœ๋ชจ ๋ณด์•ˆ ์ด๋‹ˆ์…”ํ‹ฐ๋ธŒ์™€ ๋งˆ์ฐฌ๊ฐ€์ง€๋กœ, ํŠน์ • ์กฐ์ง์ด๋‚˜ ๋ถ€์„œ์— ๊ตญํ•œ๋œ ์ ‘๊ทผ์œผ๋กœ๋Š” ์„ฑ๊ณตํ•  ์ˆ˜ ์—†๋‹ค๊ณ  ๊ฒฝ๊ณ ํ–ˆ๋‹ค.

๊ทธ๋Š” โ€œํ˜„์žฌ์™€ ๋ฏธ๋ž˜์˜ ์„ฑ๊ณต์„ ๋ณด์žฅํ•˜๋Š” ์‹คํ–‰ ๋ฐฉ์‹์„ ์ •๋ฆฝํ•˜๊ธฐ ์œ„ํ•ด์„œ๋Š” ์ „์‚ฌ ๋ชจ๋“  ๋ถ€์„œ์™€์˜ ํ˜‘์—…์ด ํ•„์š”ํ•˜๋‹คโ€๋ผ๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.

5. ๋ณด์•ˆ ์šด์˜ ๊ฐ•ํ™”๋ฅผ ์œ„ํ•œ AI ์šฐ์„  ์ „๋žต

์„ธ์ผ์ฆˆ ์„ฑ๊ณผ ๊ด€๋ฆฌ ๊ธฐ์—… ์žญํ‹€๋ฆฌ(Xactly)์˜ CISO ๋งค์Šˆ ์ƒคํ”„๋Š” ์ˆ˜์น˜ ๋ถ„์„ ๊ฒฐ๊ณผ์™€ ์œ„ํ˜‘ ํ™˜๊ฒฝ ๋ณ€ํ™” ๋ชจ๋‘๊ฐ€ AI ์‹ ๋ขฐ๋ฅผ ์ตœ์šฐ์„  ๊ณผ์ œ๋กœ ์‚ผ์•„์•ผ ํ•จ์„ ๋ณด์—ฌ์ฃผ๊ณ  ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” ์ž์‚ฌ ๋ณด์•ˆ ์šด์˜์„ ๋Œ€์ƒ์œผ๋กœ ํฌ๋ฆฌ์Šคํ…์Šจ์‹ ๋ถ„์„์„ ์ˆ˜ํ–‰ํ•œ ๊ฒฐ๊ณผ, ์ฆ์  ์ˆ˜์ง‘๊ณผ ๊ฒฝ๋ณด ๊ฒ€์ฆ, ๊ทœ์ œ ์ค€์ˆ˜ ๋ณด๊ณ ์™€ ๊ฐ™์€ ์—…๋ฌด๋ฅผ ํฌํ•จํ•ด ์ „์ฒด ๊ธฐ๋Šฅ์  ์ž‘์—…์˜ ์•ฝ 67%๊ฐ€ ๊ธฐ๊ณ„์  ์„ฑ๊ฒฉ์„ ๋ ๋ฉฐ ์ž๋™ํ™”๊ฐ€ ๊ฐ€๋Šฅํ•˜๋‹ค๋Š” ์ ์„ ํ™•์ธํ–ˆ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค.

์ƒคํ”„๋Š” โ€œ๊ณต๊ฒฉ์ž๋“ค์€ ์ด๋ฏธ ๋จธ์‹  ์†๋„๋กœ AI๋ฅผ ํ™œ์šฉํ•ด ๊ณต๊ฒฉํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ์ธ๊ฐ„์˜ ์†๋„๋กœ ๋Œ€์‘ํ•˜๋Š” ๋ฐฉ์‹์œผ๋กœ๋Š” AI ๊ธฐ๋ฐ˜ ๊ณต๊ฒฉ์„ ๋ฐฉ์–ดํ•  ์ˆ˜ ์—†๋‹คโ€๋ผ๊ณ  ์ง€์ ํ–ˆ๋‹ค. ์ด์–ด โ€œAI ์‹ ๋ขฐ๋ฅผ ๋ณด์•ˆ ์šด์˜์— ์ ์šฉํ•˜๋ฉด, ๊ธฐ๊ณ„๊ฐ€ ๋” ํšจ์œจ์ ์œผ๋กœ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ๋Š” ์ž‘์—…์„ ์ธ๊ฐ„ ๋ถ„์„๊ฐ€๊ฐ€ ์ฒ˜๋ฆฌํ•ด์•ผ ํ•˜๋Š” ์ƒํ™ฉ์„ ํ”ผํ•  ์ˆ˜ ์žˆ๊ณ , ๋™์ผํ•œ ๋ฐฉ์‹์œผ๋กœ ๋Œ€์‘ ์—ญ๋Ÿ‰์„ ๋Œ์–ด์˜ฌ๋ฆด ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

AI๊ฐ€ ๋ฐฉ์–ด ์ˆ˜๋‹จ์œผ๋กœ์„œ ์‹ค์งˆ์ ์ธ ๋„๊ตฌ๋กœ ์ž๋ฆฌ ์žก์œผ๋ฉด์„œ, CISO๋“ค์€ ์กฐ์ง ๋‚ด ๋ณด์•ˆ ํŒ€ ์šด์˜ ๋ฐฉ์‹ ์—ญ์‹œ AI์˜ ์ž ์žฌ๋ ฅ์„ ์ตœ๋Œ€ํ•œ ํ™œ์šฉํ•  ์ˆ˜ ์žˆ๋„๋ก ์žฌ๊ฒ€ํ† ํ•˜๊ณ  ์žˆ๋‹ค.

6. ๊ธฐ๋ณธ๊ฐ’์œผ๋กœ์„œ์˜ ์ œ๋กœ ํŠธ๋Ÿฌ์ŠคํŠธ ๋ชจ๋ธ ์ „ํ™˜

์†Œํ”„ํŠธ์›จ์–ด ๊ฐœ๋ฐœ ๊ธฐ์—… ์œ ๋ฆฌ์Šคํ‹ฑ(Euristiq)์˜ CTO ํŒŒ๋ธ”๋กœ ํŠธํžˆ๋ฅด๋Š” 2026๋…„ ํ•ต์‹ฌ ํ”„๋กœ์ ํŠธ๋กœ ์ž์‚ฌ ๋‚ด๋ถ€ ๊ฐœ๋ฐœ๊ณผ ๊ณ ๊ฐ ๊ฐœ๋ฐœ ์ „๋ฐ˜์— ์ œ๋กœ ํŠธ๋Ÿฌ์ŠคํŠธ ์•„ํ‚คํ…์ฒ˜๋ฅผ ๊ตฌํ˜„ํ•˜๋Š” ์ž‘์—…์„ ๊ผฝ์•˜๋‹ค. ๊ทธ๋Š” โ€œ๋ณด์•ˆ์ด ์ค‘์š”ํ•œ ๊ธฐ์—…๊ณผ ์˜ค๋žซ๋™์•ˆ ํ˜‘๋ ฅํ•ด ์™”์ง€๋งŒ, 2026๋…„์—๋Š” ์‹œ์žฅ๊ณผ ๊ทœ์ œ ์š”๊ตฌ ์ˆ˜์ค€์ด ํฌ๊ฒŒ ๋†’์•„์ง€๋ฉด์„œ โ€˜์ œ๋กœ ํŠธ๋Ÿฌ์ŠคํŠธ ๊ธฐ๋ณธ๊ฐ’โ€™ ๋ชจ๋ธ๋กœ์˜ ์ „ํ™˜์ด ์ „๋žต์  ํ•„์ˆ˜ ๊ณผ์ œ๊ฐ€ ๋  ๊ฒƒโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

ํŠธํžˆ๋ฅด๋Š” ์ด ํ”„๋กœ์ ํŠธ๊ฐ€ ๋‹จ์ˆœํžˆ ์ž์‚ฌ ๋ณด์•ˆ์„ ๊ฐ•ํ™”ํ•˜๋Š” ๋ฐ ๊ทธ์น˜์ง€ ์•Š๋Š”๋‹ค๊ณ  ๋ฐํ˜”๋‹ค. ๊ทธ๋Š” โ€œ๊ณ ๋ถ€ํ•˜ ์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ์‹œ์Šคํ…œ๋ถ€ํ„ฐ ๋ฐ์ดํ„ฐ ๋ฌด๊ฒฐ์„ฑ์ด ์ค‘์š”ํ•œ AI ๊ธฐ๋ฐ˜ ์†”๋ฃจ์…˜์— ์ด๋ฅด๊ธฐ๊นŒ์ง€, ๊ณ ๊ฐ์„ ์œ„ํ•œ ๋ณด๋‹ค ์•ˆ์ „ํ•œ ํ”Œ๋žซํผ์„ ๊ตฌ์ถ•ํ•  ์ˆ˜ ์žˆ๊ฒŒ ๋  ๊ฒƒโ€์ด๋ผ๋ฉฐ โ€œ์ธํ”„๋ผ์™€ ๊ฐœ๋ฐœ, CI/CD, ๋‚ด๋ถ€ ๋„๊ตฌ ์ „๋ฐ˜์— ์ œ๋กœ ํŠธ๋Ÿฌ์ŠคํŠธ๋ฅผ ์ ์šฉํ•จ์œผ๋กœ์จ ํ†ตํ•ฉ๋œ ๋ณด์•ˆ ๊ธฐ์ค€์„ ๋งˆ๋ จํ•˜๊ณ , ์ด๋ฅผ ๊ณ ๊ฐ ์•„ํ‚คํ…์ฒ˜์—๋„ ์ด์ „ํ•  ๊ณ„ํšโ€์ด๋ผ๊ณ  ์ „ํ–ˆ๋‹ค.

์ด ์ด๋‹ˆ์…”ํ‹ฐ๋ธŒ๋Š” ํŠน์ • ๋ณด์•ˆ ์‚ฌ๊ณ ๋ฅผ ๊ณ„๊ธฐ๋กœ ์‹œ์ž‘๋œ ๊ฒƒ์€ ์•„๋‹ˆ๋ผ๊ณ  ํŠธํžˆ๋ฅด๋Š” ์„ค๋ช…ํ–ˆ๋‹ค. ๊ทธ๋Š” โ€œ์œ„ํ˜‘ ๋ชจ๋ธ์ด ๊ทธ ์–ด๋А ๋•Œ๋ณด๋‹ค ๋น ๋ฅด๊ฒŒ ๋ณ€ํ™”ํ•˜๊ณ  ์žˆ๋‹ค๋Š” ์ ์„ ํ™•์ธํ–ˆ๋‹คโ€๋ผ๋ฉฐ โ€œ๊ณต๊ฒฉ์€ ๋” ์ด์ƒ ๊ฒฝ๊ณ„์—์„œ๋งŒ ๋ฐœ์ƒํ•˜๋Š” ๊ฒƒ์ด ์•„๋‹ˆ๋ผ, ๋ผ์ด๋ธŒ๋Ÿฌ๋ฆฌ ์ทจ์•ฝ์ ์ด๋‚˜ API, ์ทจ์•ฝํ•œ ์ธ์ฆ ๋ฉ”์ปค๋‹ˆ์ฆ˜, ์ž˜๋ชป ์„ค์ •๋œ ๊ถŒํ•œ ๋“ฑ ๋‚ด๋ถ€ ์š”์†Œ๋ฅผ ํ†ตํ•ด ์ ์  ๋” ๋งŽ์ด ๋ฐœ์ƒํ•˜๊ณ  ์žˆ๋‹คโ€๋ผ๊ณ  ๋ถ„์„ํ–ˆ๋‹ค. ์ด๋Ÿฌํ•œ ๋ณ€ํ™”๊ฐ€ ์ ‘๊ทผ ๋ฐฉ์‹์„ ๊ทผ๋ณธ์ ์œผ๋กœ ์žฌ๊ฒ€ํ† ํ•˜๊ฒŒ ๋งŒ๋“  ๊ณ„๊ธฐ๋ผ๊ณ  ๋ฐํ˜”๋‹ค.

7. ์ „์‚ฌ ์ฐจ์›์˜ ๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค ๊ฐ•ํ™”

์—”ํ„ฐํ”„๋ผ์ด์ฆˆ ๋ฐ์ดํ„ฐยทAIยท๋ฐ์ดํ„ฐ ํŒจ๋ธŒ๋ฆญ ์†”๋ฃจ์…˜ ๊ธฐ์—… ์†”๋ฆญ์Šค ํ…Œํฌ๋†€๋กœ์ง€์Šค(Solix Technologies)์˜ ์ด๊ด„ ๋ฐฐ๋ฆฌ ์ฟค์ŠคํŠธ๋Š” 2026๋…„ ์šฐ์„  ๊ณผ์ œ๋กœ ๋ชจ๋“  ์ „์‚ฌ ์‹œ์Šคํ…œ์— ๊ฑธ์นœ ํ†ตํ•ฉ ๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค ๋ฐ ๋ณด์•ˆ ํ”„๋ ˆ์ž„์›Œํฌ ๊ตฌ์ถ•์„ ์ œ์‹œํ–ˆ๋‹ค. ๊ทธ๋Š” ์ด ์ด๋‹ˆ์…”ํ‹ฐ๋ธŒ๊ฐ€ ๋งŽ์€ ์กฐ์ง์ด ์—ฌ์ „ํžˆ ๊ฒช๊ณ  ์žˆ๋Š” ์„€๋„ ๋ฐ์ดํ„ฐ ๋ฌธ์ œ์™€ ์ผ๊ด€๋˜์ง€ ์•Š์€ ์ ‘๊ทผ ํ†ต์ œ, ๊ทœ์ œ ์ค€์ˆ˜ ๊ณต๋ฐฑ์„ ํ•ด์†Œํ•˜๊ธฐ ์œ„ํ•œ ๋ชฉ์ ๋„ ํ•จ๊ป˜ ๋‹ด๊ณ  ์žˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์ฟค์ŠคํŠธ๋Š” โ€œ๋ชจ๋“  ํ™˜๊ฒฝ์—์„œ ๋ฐ์ดํ„ฐ ๋ถ„๋ฅ˜์™€ ๋ณดํ˜ธ, ๋ชจ๋‹ˆํ„ฐ๋ง ๋ฐฉ์‹์„ ํ‘œ์ค€ํ™”ํ•˜๋ฉด ์ถ”์ ๋˜์ง€ ์•Š๋Š” ๋ฏผ๊ฐ ๋ฐ์ดํ„ฐ๋ผ๋Š” ๊ฐ€์žฅ ํฐ ๋ณด์•ˆ ํ—ˆ์ ์„ ์ค„์ผ ์ˆ˜ ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ์ด ํ”„๋กœ์ ํŠธ๋Š” ๊ฐ€์‹œ์„ฑ์„ ๋†’์ด๊ณ  ์ •์ฑ… ๊ธฐ๋ฐ˜ ํ†ต์ œ๋ฅผ ๊ฐ•ํ™”ํ•ด ๋ฉ€ํ‹ฐํด๋ผ์šฐ๋“œ ํ™˜๊ฒฝ์—์„œ์˜ ๋…ธ์ถœ์„ ์ค„์ด๋Š” ๋ฐฉ์‹์œผ๋กœ ๋ณด์•ˆ ์ˆ˜์ค€์„ ๋Œ์–ด์˜ฌ๋ฆด ๊ฒƒโ€์ด๋ผ๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.

๊ทธ๋Š” ๊ณ ๊ฐ๋“ค์ด ๊ธ‰๊ฒฉํ•œ ๋ฐ์ดํ„ฐ ์ฆ๊ฐ€์™€ ์ƒˆ๋กœ์šด ๊ทœ์ œ ์š”๊ตฌ์— ์••๋„๋˜๋Š” ์ƒํ™ฉ์„ ๋ชฉ๊ฒฉํ•œ ์ดํ›„ ์ด๋ฒˆ ํ”„๋กœ์ ํŠธ๋ฅผ ์ถ”์ง„ํ•˜๊ฒŒ ๋๋‹ค๊ณ  ๋ฐํ˜”๋‹ค. ํ˜„์žฌ ๋ณด์•ˆ ๋ฐ ํด๋ผ์šฐ๋“œ ์—”์ง€๋‹ˆ์–ด๋ง ํŒ€์ด ์ฃผ์š” ๊ธฐ์ˆ  ํŒŒํŠธ๋„ˆ์™€ ํ˜‘๋ ฅํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, 2026๋…„ 3๋ถ„๊ธฐ ๋„์ž…์„ ๋ชฉํ‘œ๋กœ ์ค€๋น„๊ฐ€ ์ง„ํ–‰ ์ค‘์ด๋ผ๊ณ  ์ „ํ–ˆ๋‹ค.
dl-ciokorea@foundryco.com

When it comes to AI, not all data is created equal

14 January 2026 at 05:00

Gen AI is becoming a disruptive influence on nearly every industry, but using the best AI models and tools isnโ€™t enough. Everybodyโ€™s using the same ones but what really creates competitive advantage is being able to train and fine-tune your own models, or provide unique context to them, and that requires data.

Your companyโ€™s extensive code base, documentation, and change logs? Thatโ€™s data for your coding agents. Your library of past proposals and contracts? Data for your writing assistants. Your customer databases and support tickets? Data for your customer service chatbot.

But just because all this data exists, doesnโ€™t mean itโ€™s good.

โ€œItโ€™s so easy to point your models to any data thatโ€™s available,โ€ says Manju Naglapur, SVP and GM of cloud, applications, and infrastructure solutions at Unisys. โ€œFor the past three years, weโ€™ve seen this mistake made over and over again. The old adage garbage in, garbage out still holds true.โ€

According to a Boston Consulting Group survey released in September, 68% of 1,250 senior AI decision makers said the lack of access to high-quality data was a key challenge when it came to adopting AI. Other recent research confirms this. In an October Cisco survey of over 8,000 AI leaders, only 35% of companies have clean, centralized data with real-time integration for AI agents. And by 2027, according to IDC, companies that donโ€™t prioritize high-quality, AI-ready data will struggle scaling gen AI and agentic solutions, resulting in a 15% productivity loss.

Losing track of the semantics

Another problem using data thatโ€™s all lumped together is that the semantic layer gets confused. When data comes from multiple sources, the same type of information can be defined and structured in many ways. And as the number of data sources proliferates due to new projects or new acquisitions, the challenge increases. Even just keeping track of customers โ€” the most critical data type โ€” and basic data issues are difficult for many companies.

Dun & Bradstreet reported last year that more than half of organizations surveyed have concerns about the trustworthiness and quality of the data theyโ€™re leveraging for AI. For example, in the financial services sector, 52% of companies say AI projects have failed because of poor data. And for 44%, data quality is their biggest concern for 2026, second only to cybersecurity, based on a survey of over 2,000 industry professionals released in December.

Having multiple conflicting data standards is a challenge for everybody, says Eamonn Oโ€™Neill, CTO at Lemongrass, a cloud consultancy.

โ€œEvery mismatch is a risk,โ€ he says. โ€œBut humans figure out ways around it.โ€

AI can also be configured to do something similar, he adds, if you understand what the challenge is, and dedicate time and effort to address it. Even if the data is clean, a company should still go through a semantic mapping exercise. And if the data isnโ€™t perfect, itโ€™ll take time to tidy it up.

โ€œTake a use case with a small amount of data and get it right,โ€ he says. โ€œThatโ€™s feasible. And then you expand. Thatโ€™s what successful adoption looks like.โ€

Unmanaged and unstructured

Another mistake companies make when connecting AI to company information is to point AI at unstructured data sources, says Oโ€™Neill. And, yes, LLMs are very good at reading unstructured data and making sense of text and images. The problem is not all documents are worthy of the AIโ€™s attention.

Documents could be out of date, for example. Or they could be early versions of documents that havenโ€™t been edited yet, or that have mistakes in them.

โ€œPeople see this all the time,โ€ he says. โ€œWe connect your OneDrive or your file storage to a chatbot, and suddenly it canโ€™t tell the difference between โ€˜version 2โ€™ and โ€˜version 2 final.โ€™โ€

Itโ€™s very difficult for human users to maintain proper version control, he adds. โ€œMicrosoft can handle the different versions for you, but people still do โ€˜save asโ€™ and you end up with a plethora of unstructured data,โ€ Oโ€™Neill says.

Losing track of security

When CIOs typically think of security as it relates to AI systems, they might consider guardrails on the models, or protections around the training data and the data used for RAG embeddings. But as chatbot-based AI evolves into agentic AI, the security problems get more complex.

Say for example thereโ€™s a database of employee salaries. If an employee has a question about their salary and asks an AI chatbot embedded into their AI portal, the RAG embedding approach would be to collect only the relevant data from the database using traditional code, embed it into the prompt, then send the query off to the AI. The AI only sees the information itโ€™s allowed to see and the traditional, deterministic software stack handles the problem of keeping the rest of the employee data secure.

But when the system evolves into an agentic one, the AI agents can query the databases autonomously via MCP servers, and since they need to be able to answer questions from any employee, they require access to all employee data, and keeping it from getting into the wrong hands becomes a big task.

According to the Cisco survey, only 27% of companies have dynamic and detailed access controls for AI systems, and fewer than half feel confident in safeguarding sensitive data or preventing unauthorized access.

And the situation gets even more complicated if all the data is collected into a data lake, says Oโ€™Neill.

โ€œIf youโ€™ve put in data from lots of different sources, each of those individual sources might have its own security model,โ€ he says. โ€œWhen you pile it all into block storage, you lose that granularity of control.โ€

Trying to add the security layer in after the fact can be difficult. The solution, he says, is to go directly to the original data sources and skip the data lake entirely.

โ€œIt was about keeping history forever because storage was so cheap, and machine learning could see patterns over time and trends,โ€ he says. โ€œPlus, cross-disciplinary patterns could be spotted if you mix data from different sources.โ€

In general, data access changes dramatically when instead of humans, AI agents are involved, says Doug Gilbert, CIO and CDO at Sutherland Global, a digital transformation consultancy.

โ€œWith humans, thereโ€™s a tremendous amount of security that lives around the human,โ€ he says. โ€œFor example, most user interfaces have been written so if itโ€™s a number-only field, you canโ€™t put a letter in there. But once you put in an AI, all thatโ€™s gone. Itโ€™s a raw back door into your systems.โ€

The speed trap

But the number-one mistake Gilbert sees CIOs making is they simply move too fast. โ€œThis is why most projects fail,โ€ he says. โ€œThereโ€™s such a race for speed.โ€

Too often, CIOs look at data issues as slowdowns, but all those things are massive risks, he adds. โ€œA lot of people doing AI projects are going to get audited and theyโ€™ll have to stop and re-do everything,โ€ he says.

So getting the data right isnโ€™t a slowdown. โ€œWhen you put the proper infrastructure in place, then you speed through your innovation, you pass audits, and you have compliance,โ€ he says.

Another area that might feel like an unnecessary waste of time is testing. Itโ€™s not always a good strategy to move fast, break things, and then fix them later on after deployment.

โ€œWhatโ€™s the cost of a mistake that moves at the speed of light?โ€ he asks. โ€œI would always go to testing first. Itโ€™s amazing how many products we see that are pushed to market without any testing.โ€

Putting AI to work to fix the data

The lack of quality data might feel like a hopeless problem thatโ€™s only going to get worse as AI use cases expand.

In an October AvePoint report based on a survey of 775 global business leaders, 81% of organizations have already delayed deployment of AI assistants due to data management or data security issues, with an average delay of six months.

Meanwhile, not only the number of AI projects continues to grow but also the amount of data. Nearly 52% of respondents also said their companies were managing more than 500 petabytes of data, up from just 41% a year ago.

But Unisysโ€™ Naglapur says itโ€™s going to become easier to get a 360-degree view of a customer, and to clean up and reconcile other data sources, because of AI.

โ€œThis is the paradox,โ€ he says. โ€œAI will help with everything. If you think about a digital transformation that would take three years, you can do it now in 12 to 18 months with AI.โ€ The tools are getting closer to reality, and theyโ€™ll accelerate the pace of change, he says.

AI๊ฐ€ ๋ฐ”๊พผ ๊ณต๊ฒฉ ์ง€ํ˜•ยทยทยท๊ธฐ์—… ์ธํ”„๋ผ๋ฅผ ๊ฒจ๋ƒฅํ•œ ์‹ ์ข… ์œ„ํ˜‘๊ณผ ๋ณด์•ˆ ์ „๋žต

9 January 2026 at 01:15

AI ์‹œ์Šคํ…œ๊ณผ ์ธํ”„๋ผ๋ฅผ ๊ฒจ๋ƒฅํ•œ ๊ณต๊ฒฉ์€ ์‹ค์ œ ์‚ฌ๋ก€๋ฅผ ํ†ตํ•ด ์ ์ฐจ ๊ตฌ์ฒด์ ์ธ ํ˜•ํƒœ๋ฅผ ๋ ๊ณ  ์žˆ์œผ๋ฉฐ, ๋ณด์•ˆ ์ „๋ฌธ๊ฐ€๋“ค์€ ํ–ฅํ›„ ๋ช‡ ๋…„ ์•ˆ์— ์ด๋Ÿฌํ•œ ๊ณต๊ฒฉ ์œ ํ˜•์ด ๋”์šฑ ์ฆ๊ฐ€ํ•  ๊ฒƒ์œผ๋กœ ๋‚ด๋‹ค๋ณด๊ณ  ์žˆ๋‹ค. AI ๋„์ž…์„ ํ†ตํ•ด ๋น ๋ฅด๊ฒŒ ์„ฑ๊ณผ๋ฅผ ๋‚ด๋ ค๋Š” ๊ณผ์ •์—์„œ, ๋Œ€๋ถ€๋ถ„์˜ ์กฐ์ง์€ AI ๋„๊ตฌ์™€ ํ™œ์šฉ ์‚ฌ๋ก€๋ฅผ ๋ฐฐํฌํ•˜๋ฉด์„œ ๋ณด์•ˆ ๊ฐ•ํ™”๋ฅผ ๋А์Šจํ•˜๊ฒŒ ์ ์šฉํ•ด ์™”๋‹ค. ๊ทธ ๊ฒฐ๊ณผ, ์ƒ๋‹น์ˆ˜ ์กฐ์ง์ด ์ด๋Ÿฌํ•œ ๊ณต๊ฒฉ์„ ํƒ์ง€ํ•˜๊ฑฐ๋‚˜ ์ฐจ๋‹จํ•˜๊ณ , ์‚ฌ๊ณ  ๋ฐœ์ƒ ์‹œ ๋Œ€์‘ํ•  ์ค€๋น„๊ฐ€ ๋˜์–ด ์žˆ์ง€ ์•Š๋‹ค๋Š” ๊ฒฝ๊ณ ๋„ ๋‚˜์˜จ๋‹ค.

๋ฏธ๊ตญ ์‚ฌ์šฐ์Šคํ”Œ๋กœ๋ฆฌ๋‹ค๋Œ€ํ•™๊ต ๋ฒจ๋ฆฌ๋‹ˆ ์ธ๊ณต์ง€๋Šฅยท์‚ฌ์ด๋ฒ„๋ณด์•ˆยท์ปดํ“จํŒ… ์นผ๋ฆฌ์ง€์˜ ๋ถ€๊ต์ˆ˜์ด์ž ์Šคํƒ€ํŠธ์—… ์•ก์ถ”์–ผ๋ผ์ด์ œ์ด์…˜AI์˜ ๋Œ€ํ‘œ์ธ ์กด ๋ฆฌ์นดํ† ๋Š” โ€œ๋Œ€๋ถ€๋ถ„ ์ด๋Ÿฐ ๊ณต๊ฒฉ์ด ๊ฐ€๋Šฅํ•˜๋‹ค๋Š” ์ ์€ ์ธ์‹ํ•˜๊ณ  ์žˆ์ง€๋งŒ, ์œ„ํ—˜์„ ์–ด๋–ป๊ฒŒ ์ ์ ˆํžˆ ์™„ํ™”ํ•ด์•ผ ํ•˜๋Š”์ง€๊นŒ์ง€ ์ถฉ๋ถ„ํžˆ ์ดํ•ดํ•˜๊ณ  ์žˆ๋Š” ๊ฒฝ์šฐ๋Š” ๋งŽ์ง€ ์•Š๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

AI ์‹œ์Šคํ…œ์„ ๊ฒจ๋ƒฅํ•œ ์ฃผ์š” ์œ„ํ˜‘

AI ์‹œ์Šคํ…œ์„ ๋Œ€์ƒ์œผ๋กœ ํ•œ ๊ณต๊ฒฉ ์œ ํ˜•์ด ๋‹ค์–‘ํ•˜๊ฒŒ ๋“ฑ์žฅํ•˜๊ณ  ์žˆ๋‹ค. ๋ฐ์ดํ„ฐ ์ค‘๋…์ฒ˜๋Ÿผ ํ•™์Šต ๋‹จ๊ณ„์—์„œ ๋ฐœ์ƒํ•˜๋Š” ๊ณต๊ฒฉ์ด ์žˆ๋Š”๊ฐ€ ํ•˜๋ฉด, ์ ๋Œ€์  ์ž…๋ ฅ๊ณผ ๊ฐ™์ด ์ถ”๋ก  ๋‹จ๊ณ„์—์„œ ์ด๋ค„์ง€๋Š” ๊ณต๊ฒฉ๋„ ์žˆ๋‹ค. ์ด ๋ฐ–์—๋„ ๋ชจ๋ธ ํƒˆ์ทจ์ฒ˜๋Ÿผ ๋ฐฐํฌ ๋‹จ๊ณ„์—์„œ ๋ฐœ์ƒํ•˜๋Š” ๊ณต๊ฒฉ ์œ ํ˜•๋„ ์กด์žฌํ•œ๋‹ค.

๋‹ค์Œ์€ ํ˜„์žฌ AI ์ธํ”„๋ผ์™€ ๊ด€๋ จํ•ด ์ „๋ฌธ๊ฐ€๋“ค์ด ๊ฒฝ๊ณ ํ•˜๋Š” ์ฃผ์š” ์œ„ํ˜‘ ์œ ํ˜•์„ ์ •๋ฆฌํ•œ ๊ฒƒ์ด๋‹ค. ์ผ๋ถ€๋Š” ๋น„๊ต์  ๋“œ๋ฌผ๊ฑฐ๋‚˜ ์ด๋ก ์  ๋‹จ๊ณ„์— ๋จธ๋ฌผ๋Ÿฌ ์žˆ์ง€๋งŒ, ์ƒ๋‹น์ˆ˜๋Š” ์‹ค์ œ ํ™˜๊ฒฝ์—์„œ ๊ด€์ธก๋๊ฑฐ๋‚˜ ์—ฐ๊ตฌ์ž๋“ค์ด ์ฃผ๋ชฉํ•  ๋งŒํ•œ ๊ฐœ๋… ์ฆ๋ช…์„ ํ†ตํ•ด ๊ฐ€๋Šฅ์„ฑ์„ ์ž…์ฆํ•œ ์‚ฌ๋ก€๋‹ค.

๋ฐ์ดํ„ฐ ์ค‘๋…

๋ฐ์ดํ„ฐ ์ค‘๋…(Data Poisoning)์€ ์•…์˜์ ์ธ ๊ณต๊ฒฉ์ž๊ฐ€ ๋จธ์‹ ๋Ÿฌ๋‹ ๋ชจ๋ธ์„ ํฌํ•จํ•œ AI ์‹œ์Šคํ…œ์„ ๊ฐœ๋ฐœํ•˜๊ฑฐ๋‚˜ ํ•™์Šตํ•˜๋Š” ๋ฐ ์‚ฌ์šฉ๋˜๋Š” ๋ฐ์ดํ„ฐ๋ฅผ ์กฐ์ž‘ยท๋ณ€์กฐยท์˜ค์—ผ์‹œํ‚ค๋Š” ๊ณต๊ฒฉ ์œ ํ˜•์ด๋‹ค. ๋ฐ์ดํ„ฐ๋ฅผ ์†์ƒ์‹œํ‚ค๊ฑฐ๋‚˜ ์ž˜๋ชป๋œ ๋ฐ์ดํ„ฐ๋ฅผ ์ฃผ์ž…ํ•จ์œผ๋กœ์จ ๋ชจ๋ธ์˜ ์„ฑ๋Šฅ์„ ์™œ๊ณกํ•˜๊ฑฐ๋‚˜ ํŽธํ–ฅ์‹œํ‚ค๊ณ , ๊ฒฐ๊ณผ์ ์œผ๋กœ ๋ถ€์ •ํ™•ํ•œ ํŒ๋‹จ์„ ๋‚ด๋ฆฌ๋„๋ก ๋งŒ๋“ค ์ˆ˜ ์žˆ๋‹ค.

๋ณด์•ˆ ๊ต์œก ๋ฐ ์ž๊ฒฉ ์ธ์ฆ ๊ธฐ๊ด€ ์ƒŒ์Šค(SANS)์˜ ์ตœ๊ณ  AI ์ฑ…์ž„์ž(CAIO) ๊ฒธ ์—ฐ๊ตฌ ์ด๊ด„์ธ ๋กœ๋ฒ„ํŠธ T. ๋ฆฌ๋Š” โ€œ์˜ˆ๋ฅผ ๋“ค์–ด, ์ดˆ๋ก์ƒ‰์€ โ€˜๊ฐ€๋ผโ€™๊ฐ€ ์•„๋‹ˆ๋ผ โ€˜๋ฉˆ์ถฐ๋ผโ€™๋ผ๋Š” ์˜๋ฏธ๋ผ๊ณ  ๋ชจ๋ธ์— ํ•™์Šต์‹œํ‚ค๋Š” ๊ณต๊ฒฉ์„ ์ƒ์ƒํ•ด๋ณผ ์ˆ˜ ์žˆ๋‹คโ€๋ผ๋ฉฐ โ€œ๋ชจ๋ธ์˜ ์ถœ๋ ฅ ๊ฒฐ๊ณผ ์ž์ฒด๋ฅผ ์ €ํ•˜์‹œํ‚ค๋Š” ๊ฒƒ์ด ๋ชฉ์ โ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ชจ๋ธ ์ค‘๋…

๋ชจ๋ธ ์ค‘๋…(Model poisoning)์€ ๋ฐ์ดํ„ฐ๊ฐ€ ์•„๋‹ˆ๋ผ ๋ชจ๋ธ ์ž์ฒด๋ฅผ ์ง์ ‘ ๊ณต๊ฒฉ ๋Œ€์ƒ์œผ๋กœ ์‚ผ๋Š”๋‹ค. ๋ชจ๋ธ์˜ ๊ตฌ์กฐ๋‚˜ ํŒŒ๋ผ๋ฏธํ„ฐ๋ฅผ ๋ณ€์กฐํ•ด ๋ถ€์ •ํ™•ํ•œ ๊ฒฐ๊ณผ๊ฐ€ ๋‚˜์˜ค๋„๋ก ์œ ๋„ํ•˜๋Š” ๋ฐฉ์‹์ด๋‹ค. ์ผ๋ถ€ ์ •์˜์—์„œ๋Š” ๋ฐ์ดํ„ฐ ์ค‘๋…์œผ๋กœ ์ธํ•ด ํ•™์Šต ๋ฐ์ดํ„ฐ๊ฐ€ ์†์ƒ๋œ ๊ฒฝ์šฐ๊นŒ์ง€ ๋ชจ๋ธ ์ค‘๋…์˜ ๋ฒ”์ฃผ์— ํฌํ•จํ•˜๊ธฐ๋„ ํ•œ๋‹ค.

ํˆด ์ค‘๋…

ํˆด ์ค‘๋…(Tool poisoning)์€ 2025๋…„ ๋ด„ ์ธ๋ฒ ๋ฆฌ์–ธํŠธ ๋žฉ์Šค(Invariant Labs)๊ฐ€ ์‹๋ณ„ํ•œ ๊ณต๊ฒฉ ์œ ํ˜•์ด๋‹ค. ์ธ๋ฒ ๋ฆฌ์–ธํŠธ๋Š” ์ž์ฒด ์—ฐ๊ตฌ ๊ฒฐ๊ณผ๋ฅผ ๊ณต์œ ํ•˜๋ฉฐ โ€œํˆด ์ค‘๋… ๊ณต๊ฒฉ์„ ๊ฐ€๋Šฅํ•˜๊ฒŒ ํ•˜๋Š” ๋ชจ๋ธ ์ปจํ…์ŠคํŠธ ํ”„๋กœํ† ์ฝœ(MCP)์˜ ์‹ฌ๊ฐํ•œ ์ทจ์•ฝ์ ์„ ๋ฐœ๊ฒฌํ–ˆ๋‹คโ€๋ผ๋ฉฐ โ€œ์ด ์ทจ์•ฝ์ ์€ ๋ฏผ๊ฐํ•œ ๋ฐ์ดํ„ฐ ์œ ์ถœ๊ณผ AI ๋ชจ๋ธ์˜ ๋น„์ธ๊ฐ€ ํ–‰์œ„๋กœ ์ด์–ด์งˆ ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ๋ฐํ˜”๋‹ค.

๋˜ํ•œ ์‹คํ—˜ ๊ฒฐ๊ณผ๋ฅผ ํ†ตํ•ด โ€œ์•…์„ฑ ์„œ๋ฒ„๊ฐ€ ์‚ฌ์šฉ์ž๋กœ๋ถ€ํ„ฐ ๋ฏผ๊ฐํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ์œ ์ถœํ•  ์ˆ˜ ์žˆ์„ ๋ฟ ์•„๋‹ˆ๋ผ, ์‹ ๋ขฐ๋œ ๋‹ค๋ฅธ ์„œ๋ฒ„๊ฐ€ ์ œ๊ณตํ•œ ์ง€์นจ์„ ๋ฎ์–ด์“ฐ๊ณ  ์—์ด์ „ํŠธ์˜ ํ–‰๋™์„ ํƒˆ์ทจํ•ด, ์‹ ๋ขฐ๋œ ์ธํ”„๋ผ๋ฅผ ํฌํ•จํ•œ ์—์ด์ „ํŠธ ๊ธฐ๋Šฅ ์ „๋ฐ˜์„ ์™„์ „ํžˆ ์žฅ์•…ํ•  ์ˆ˜ ์žˆ๋‹คโ€๋Š” ์ ์„ ํ™•์ธํ–ˆ๋‹ค๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์ด ๊ณต๊ฒฉ์€ MCP ํˆด ์„ค๋ช… ์•ˆ์— ์•…์„ฑ ์ง€์‹œ๋ฅผ ์‚ฝ์ž…ํ•˜๋Š” ๋ฐฉ์‹์œผ๋กœ ์ด๋ค„์ง„๋‹ค. AI ๋ชจ๋ธ์ด ์ด๋ฅผ ํ•ด์„ํ•˜๋Š” ๊ณผ์ •์—์„œ ๋ชจ๋ธ์˜ ๋™์ž‘์ด ํƒˆ์ทจ๋œ๋‹ค. ์ปจ์„คํŒ… ๊ธฐ์—… ์ปจ์Šคํ…”๋ ˆ์ด์…˜ ๋ฆฌ์„œ์น˜(Constellation Research)์˜ ๋ถ€์‚ฌ์žฅ ๊ฒธ ์ˆ˜์„ ์• ๋„๋ฆฌ์ŠคํŠธ ์น˜๋ฝ ๋ฉ”ํƒ€๋Š” โ€œ์—์ด์ „ํŠธ๊ฐ€ ํŠน์ • ํ–‰๋™์„ ํ•˜๋„๋ก ์†์ด๊ธฐ ์œ„ํ•ด MCP ๊ณ„์ธต ์ž์ฒด๋ฅผ ์˜ค์—ผ์‹œํ‚ค๋Š” ๊ณต๊ฒฉโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

ํ”„๋กฌํ”„ํŠธ ์ธ์ ์…˜

ํ”„๋กฌํ”„ํŠธ ์ธ์ ์…˜(Prompt injection) ๊ณต๊ฒฉ์€ ๊ฒ‰๋ณด๊ธฐ์—๋Š” ์ •์ƒ์ ์ธ ์š”์ฒญ์ฒ˜๋Ÿผ ๋ณด์ด์ง€๋งŒ, ์‹ค์ œ๋กœ๋Š” ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ์ด ์ˆ˜ํ–‰ํ•ด์„œ๋Š” ์•ˆ ๋˜๋Š” ํ–‰๋™์„ ํ•˜๋„๋ก ์œ ๋„ํ•˜๋Š” ์•…์„ฑ ๋ช…๋ น์„ ํฌํ•จํ•œ ํ”„๋กฌํ”„ํŠธ๋ฅผ ์‚ฌ์šฉํ•˜๋Š” ๋ฐฉ์‹์ด๋‹ค. ๊ณต๊ฒฉ์ž๋Š” ์ด๋ฅผ ํ†ตํ•ด ๋ชจ๋ธ์˜ ๊ฐ€๋“œ๋ ˆ์ผ์„ ์šฐํšŒํ•˜๊ฑฐ๋‚˜ ๋ฌด๋ ฅํ™”ํ•˜๊ณ , ๋ฏผ๊ฐํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ๋…ธ์ถœ์‹œํ‚ค๊ฑฐ๋‚˜ ๋น„์ธ๊ฐ€ ์ž‘์—…์„ ์ˆ˜ํ–‰ํ•˜๊ฒŒ ๋งŒ๋“ ๋‹ค.

๊ธ€๋กœ๋ฒŒ ์ปจ์„คํŒ… ๊ธฐ์—… ๋”์ปค ์นผ๋ผ์ผ(Ducker Carlisle)์˜ ์ตœ๊ณ  ๋ฐ์ดํ„ฐยทAI ์ฑ…์ž„์ž ํŒŒ๋น„์•™ ํฌ๋กœ์Šค๋Š” โ€œํ”„๋กฌํ”„ํŠธ ์ธ์ ์…˜์„ ํ™œ์šฉํ•˜๋ฉด AI ์—์ด์ „ํŠธ๊ฐ€ ์›๋ž˜ ์ˆ˜ํ–‰ํ•ด์•ผ ํ•  ์—ญํ•  ์ž์ฒด๋ฅผ ๋ฐ”๊ฟ€ ์ˆ˜ ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

์ตœ๊ทผ์—๋Š” ์ฑ—GPT๊ฐ€ ์Šค์Šค๋กœ ํ”„๋กฌํ”„ํŠธ ์ธ์ ์…˜์„ ํ•˜๋„๋ก ์œ ๋„ํ•œ ์‚ฌ๋ก€๋ฅผ ๋น„๋กฏํ•ด, ๋ฌธ์„œ ๋งคํฌ๋กœ์— ์•…์„ฑ ํ”„๋กฌํ”„ํŠธ๋ฅผ ์‚ฝ์ž…ํ•œ ๊ณต๊ฒฉ, ์ธ๊ธฐ AI ์—์ด์ „ํŠธ๋ฅผ ๋Œ€์ƒ์œผ๋กœ ํ•œ ์ œ๋กœ ํด๋ฆญ ํ”„๋กฌํ”„ํŠธ ๊ณต๊ฒฉ ์‹œ์—ฐ ๋“ฑ ์—ฌ๋Ÿฌ ์ฃผ๋ชฉํ•  ๋งŒํ•œ ์‚ฌ๋ก€์™€ ๊ฐœ๋… ์ฆ๋ช…์ด ์ž‡๋”ฐ๋ผ ๋ณด๊ณ ๋˜๊ณ  ์žˆ๋‹ค.

์ ๋Œ€์  ์ž…๋ ฅ

๋ชจ๋ธ ์†Œ์œ ์ž์™€ ์šด์˜์ž๋Š” ๋ณ€ํ˜•๋œ ๋ฐ์ดํ„ฐ๋ฅผ ํ™œ์šฉํ•ด ๋ชจ๋ธ์˜ ๋ณต์›๋ ฅ์„ ์ ๊ฒ€ํ•˜์ง€๋งŒ, ๊ณต๊ฒฉ์ž๋Š” ์ด๋ฅผ ๊ต๋ž€ ์ˆ˜๋‹จ์œผ๋กœ ์‚ฌ์šฉํ•œ๋‹ค. ์ ๋Œ€์  ์ž…๋ ฅ(Adversarial inputs) ๊ณต๊ฒฉ์€ ์•…์˜์ ์ธ ํ–‰์œ„์ž๊ฐ€ ๋ชจ๋ธ์— ๊ธฐ๋งŒ์ ์ธ ๋ฐ์ดํ„ฐ๋ฅผ ์ž…๋ ฅํ•ด ์ž˜๋ชป๋œ ๊ฒฐ๊ณผ๋ฅผ ์ถœ๋ ฅํ•˜๋„๋ก ์œ ๋„ํ•˜๋Š” ๋ฐฉ์‹์ด๋‹ค.

์ด๋•Œ ์ž…๋ ฅ ๋ฐ์ดํ„ฐ์˜ ๋ณ€ํ™”๋Š” ๋Œ€์ฒด๋กœ ๋งค์šฐ ๋ฏธ์„ธํ•˜๊ฑฐ๋‚˜ ๋‹จ์ˆœํ•œ ๋…ธ์ด์ฆˆ ์ˆ˜์ค€์— ๊ทธ์นœ๋‹ค. ๋ณด์•ˆ ์‹œ์Šคํ…œ์˜ ํƒ์ง€๋ฅผ ํ”ผํ•  ๋งŒํผ ๋ˆˆ์— ๋„์ง€ ์•Š๋„๋ก ์„ค๊ณ„๋˜์ง€๋งŒ, ๋™์‹œ์— ๋ชจ๋ธ์˜ ํŒ๋‹จ์„ ํํŠธ๋Ÿฌ๋œจ๋ฆด ์ˆ˜ ์žˆ์„ ์ •๋„๋กœ ํšจ๊ณผ๋ฅผ ๋ฐœํœ˜ํ•œ๋‹ค. ์ด๋Ÿฐ ํŠน์„ฑ ๋•Œ๋ฌธ์— ์ ๋Œ€์  ์ž…๋ ฅ์€ ๋Œ€ํ‘œ์ ์ธ ํšŒํ”ผ ๊ณต๊ฒฉ ์œ ํ˜•์œผ๋กœ ๋ถ„๋ฅ˜๋œ๋‹ค.

๋ชจ๋ธ ํƒˆ์ทจยท๋ชจ๋ธ ์ถ”์ถœ

์•…์˜์ ์ธ ๊ณต๊ฒฉ์ž๋Š” ๋ชจ๋ธ๊ณผ ๊ทธ ํŒŒ๋ผ๋ฏธํ„ฐ, ๋‚˜์•„๊ฐ€ ํ•™์Šต ๋ฐ์ดํ„ฐ๊นŒ์ง€ ๋ณต์ œํ•˜๊ฑฐ๋‚˜ ์—ญ์„ค๊ณ„ํ•  ์ˆ˜ ์žˆ๋‹ค. ์ผ๋ฐ˜์ ์œผ๋กœ ๊ณต๊ฐœ๋œ API, ์˜ˆ๋ฅผ ๋“ค์–ด ๋ชจ๋ธ์˜ ์˜ˆ์ธก API๋‚˜ ํด๋ผ์šฐ๋“œ ์„œ๋น„์Šค API๋ฅผ ๋ฐ˜๋ณต์ ์œผ๋กœ ํ˜ธ์ถœํ•ด ์ถœ๋ ฅ ๊ฒฐ๊ณผ๋ฅผ ์ˆ˜์ง‘ํ•˜๋Š” ๋ฐฉ์‹์ด ํ™œ์šฉ๋œ๋‹ค.

์ดํ›„ ๋ชจ๋ธ์ด ์–ด๋–ป๊ฒŒ ์‘๋‹ตํ•˜๋Š”์ง€๋ฅผ ๋ถ„์„ํ•˜๊ณ , ๊ทธ ๊ฒฐ๊ณผ๋ฅผ ๋ฐ”ํƒ•์œผ๋กœ ๋ชจ๋ธ์„ ์žฌ๊ตฌ์„ฑํ•œ๋‹ค. ๊ธ€๋กœ๋ฒŒ ์ปจ์„คํŒ… ๊ธฐ์—… PwC์˜ ๊ธ€๋กœ๋ฒŒ ์œ„ํ˜‘ ์ธํ…”๋ฆฌ์ „์Šค ์•„๋ฉ”๋ฆฌ์นด ์ด๊ด„์ธ ์•จ๋ฆฌ์Šจ ์œ„์ฝ”ํ”„๋Š” โ€œ๋„๊ตฌ ์ž์ฒด๋ฅผ ๋น„์ธ๊ฐ€ ์ƒํƒœ๋กœ ๋ณต์ œํ•  ์ˆ˜ ์žˆ๊ฒŒ ๋งŒ๋“œ๋Š” ๊ณต๊ฒฉโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ชจ๋ธ ์—ญ์ถ”๋ก 

๋ชจ๋ธ ์—ญ์ถ”๋ก ์€ ํŠน์ •ํ•œ ์ถ”์ถœ ๊ณต๊ฒฉ์˜ ํ•œ ํ˜•ํƒœ๋กœ, ๊ณต๊ฒฉ์ž๊ฐ€ ๋ชจ๋ธ ํ•™์Šต์— ์‚ฌ์šฉ๋œ ๋ฐ์ดํ„ฐ๋ฅผ ์žฌ๊ตฌ์„ฑํ•˜๊ฑฐ๋‚˜ ์ถ”๋ก ํ•˜๋ ค๋Š” ์‹œ๋„๋ฅผ ์˜๋ฏธํ•œ๋‹ค.

ํ•ด์ปค๊ฐ€ ๋ชจ๋ธ์˜ ์ถœ๋ ฅ ๊ฒฐ๊ณผ๋ฅผ ํ™œ์šฉํ•ด ํ•™์Šต์— ์‚ฌ์šฉ๋œ ์ž…๋ ฅ ์ •๋ณด๋ฅผ ๊ฑฐ๊พธ๋กœ ์ถ”์ •ํ•œ๋‹ค๋Š” ์ ์—์„œ ์ด๋Ÿฐ ์ด๋ฆ„์ด ๋ถ™์—ˆ๋‹ค. ์ด๋ฅผ ํ†ตํ•ด ํ•™์Šต ๋ฐ์ดํ„ฐ์— ํฌํ•จ๋œ ์ •๋ณด๊ฐ€ ๋…ธ์ถœ๋  ์ˆ˜ ์žˆ๋‹ค.

๊ณต๊ธ‰๋ง ์œ„ํ—˜

AI ์‹œ์Šคํ…œ์€ ์˜คํ”ˆ์†Œ์Šค ์ฝ”๋“œ์™€ ์˜คํ”ˆ์†Œ์Šค ๋ชจ๋ธ, ์„œ๋“œํŒŒํ‹ฐ ๋ชจ๋ธ, ๋‹ค์–‘ํ•œ ๋ฐ์ดํ„ฐ ์†Œ์Šค ๋“ฑ ์—ฌ๋Ÿฌ ๊ตฌ์„ฑ ์š”์†Œ๋ฅผ ๊ฒฐํ•ฉํ•ด ๋งŒ๋“ค์–ด์ง„๋‹ค. ์ด ๊ฐ€์šด๋ฐ ์–ด๋А ํ•˜๋‚˜๋ผ๋„ ๋ณด์•ˆ ์ทจ์•ฝ์ ์ด ์กด์žฌํ•˜๋ฉด, ๊ทธ ์˜ํ–ฅ์€ AI ์‹œ์Šคํ…œ ์ „๋ฐ˜์œผ๋กœ ํ™•์‚ฐ๋  ์ˆ˜ ์žˆ๋‹ค.

์ด๋กœ ์ธํ•ด AI ์‹œ์Šคํ…œ์€ ๊ณต๊ธ‰๋ง ๊ณต๊ฒฉ์— ์ทจ์•ฝํ•˜๋‹ค. ๊ณต๊ฒฉ์ž๋Š” ๊ตฌ์„ฑ ์š”์†Œ ๋‚ด๋ถ€์˜ ์ทจ์•ฝ์ ์„ ์•…์šฉํ•ด ์ „์ฒด ์‹œ์Šคํ…œ์„ ์นจํ•ดํ•  ์ˆ˜ ์žˆ๋‹ค.

ํƒˆ์˜ฅ

๋ชจ๋ธ ํƒˆ์˜ฅ(Jailbreaking)์ด๋ผ๊ณ ๋„ ๋ถˆ๋ฆฌ๋Š” ์ด ๊ณต๊ฒฉ์˜ ๋ชฉ์ ์€, ์ฃผ๋กœ ๋Œ€๊ทœ๋ชจ ์–ธ์–ด ๋ชจ๋ธ๊ณผ์˜ ์ƒํ˜ธ์ž‘์šฉ์„ ํ†ตํ•ด AI ์‹œ์Šคํ…œ์˜ ํ–‰๋™๊ณผ ์ถœ๋ ฅ์„ ์ œํ•œํ•˜๋Š” ๊ฐ€๋“œ๋ ˆ์ผ์„ ๋ฌด์‹œํ•˜๋„๋ก ๋งŒ๋“œ๋Š” ๋ฐ ์žˆ๋‹ค. ์—ฌ๊ธฐ์—๋Š” ์œ ํ•ดํ•˜๊ฑฐ๋‚˜ ๊ณต๊ฒฉ์ ์ด๊ฑฐ๋‚˜ ๋น„์œค๋ฆฌ์ ์ธ ๊ฒฐ๊ณผ๋ฅผ ๋ง‰๊ธฐ ์œ„ํ•œ ๊ฐ์ข… ๋ณดํ˜ธ ์žฅ์น˜๊ฐ€ ํฌํ•จ๋œ๋‹ค.

ํ•ด์ปค๋Š” ์ด ๊ฐ™์€ ๊ณต๊ฒฉ์„ ์ˆ˜ํ–‰ํ•˜๊ธฐ ์œ„ํ•ด ๋‹ค์–‘ํ•œ ๊ธฐ๋ฒ•์„ ํ™œ์šฉํ•œ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด ์—ญํ• ๊ทน ๊ณต๊ฒฉ์„ ์‚ฌ์šฉํ•ด, AI๊ฐ€ ๊ฐœ๋ฐœ์ž์™€ ๊ฐ™์€ ํŠน์ • ์ธ๋ฌผ์„ ์—ฐ๊ธฐํ•˜๋„๋ก ์ง€์‹œํ•จ์œผ๋กœ์จ ๊ฐ€๋“œ๋ ˆ์ผ์„ ์šฐํšŒํ•˜๊ฒŒ ๋งŒ๋“ค ์ˆ˜ ์žˆ๋‹ค. ์ •์ƒ์ ์ธ ์š”์ฒญ์ฒ˜๋Ÿผ ๋ณด์ด๋Š” ํ”„๋กฌํ”„ํŠธ ์•ˆ์— ์•…์„ฑ ์ง€์‹œ๋ฅผ ์ˆจ๊ธฐ๊ฑฐ๋‚˜, ์ธ์ฝ”๋”ฉยท์™ธ๊ตญ์–ดยทํŠน์ˆ˜ ๋ฌธ์ž ๋“ฑ์„ ํ™œ์šฉํ•ด ํ•„ํ„ฐ๋ฅผ ํšŒํ”ผํ•˜๋Š” ๋ฐฉ์‹๋„ ์žˆ๋‹ค. ๊ฐ€์ƒ์˜ ์ƒํ™ฉ์ด๋‚˜ ์—ฐ๊ตฌ ์งˆ๋ฌธ ํ˜•ํƒœ๋กœ ํ”„๋กฌํ”„ํŠธ๋ฅผ ๊ตฌ์„ฑํ•˜๊ฑฐ๋‚˜, ์—ฌ๋Ÿฌ ๋‹จ๊ณ„์˜ ์งˆ๋ฌธ์„ ํ†ตํ•ด ์ตœ์ข… ๋ชฉํ‘œ์— ๋„๋‹ฌํ•˜๋„๋ก ์œ ๋„ํ•˜๋Š” ๋ฐฉ๋ฒ•๋„ ์‚ฌ์šฉ๋œ๋‹ค.

์ด๋Ÿฌํ•œ ๊ณต๊ฒฉ์˜ ๋ชฉ์ ์€ ๋‹ค์–‘ํ•˜๋ฉฐ, AI ์‹œ์Šคํ…œ์ด ์•…์„ฑ ์ฝ”๋“œ๋ฅผ ์ž‘์„ฑํ•˜๊ฒŒ ํ•˜๊ฑฐ๋‚˜, ๋ฌธ์ œ๊ฐ€ ๋  ์ˆ˜ ์žˆ๋Š” ์ฝ˜ํ…์ธ ๋ฅผ ํ™•์‚ฐ์‹œํ‚ค๊ฑฐ๋‚˜, ๋ฏผ๊ฐํ•œ ๋ฐ์ดํ„ฐ๋ฅผ ๋…ธ์ถœํ•˜๋„๋ก ๋งŒ๋“œ๋Š” ๊ฒฝ์šฐ๊ฐ€ ํฌํ•จ๋œ๋‹ค.

๋ฆฌ์นดํ† ๋Š” โ€œ์ฑ„ํŒ… ์ธํ„ฐํŽ˜์ด์Šค๊ฐ€ ์กด์žฌํ•˜๋Š” ํ•œ, ์ด๋ฅผ ์ด์šฉํ•ด ์‹œ์Šคํ…œ์ด ์„ค์ •๋œ ๋ฒ”์œ„๋ฅผ ๋ฒ—์–ด๋‚˜๋„๋ก ์ž‘๋™ํ•˜๊ฒŒ ๋งŒ๋“œ๋Š” ๋ฐฉ์‹์€ ์กด์žฌํ•œ๋‹คโ€๋ผ๋ฉฐ โ€œ์ ์  ๋” ๊ฐ•๋ ฅํ•œ ์ถ”๋ก  ์‹œ์Šคํ…œ์„ ๊ฐ–๊ฒŒ ๋˜๋ฉด์„œ ๊ฐ์ˆ˜ํ•ด์•ผ ํ•˜๋Š” ํŠธ๋ ˆ์ด๋“œ์˜คํ”„โ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

AI ์‹œ์Šคํ…œ ์œ„ํ˜‘์— ๋Œ€์‘ํ•˜๊ธฐ

๊ฒฝ์˜์ง„์ด ์ƒ์‚ฐ์„ฑ๊ณผ ํ˜์‹ ์„ ์ด์œ ๋กœ AI ์ด๋‹ˆ์…”ํ‹ฐ๋ธŒ์— ์†๋„๋ฅผ ๋‚ด๋Š” ์ƒํ™ฉ์—์„œ, ์ตœ๊ณ ์ •๋ณด๋ณด์•ˆ์ฑ…์ž„์ž(CISO)๋Š” ํ•ด๋‹น ํ”„๋กœ์ ํŠธ์™€ ์กฐ์ง ์ „๋ฐ˜์˜ AI ์ธํ”„๋ผ ๋ณด์•ˆ์„ ์ตœ์šฐ์„  ๊ณผ์ œ๋กœ ์‚ผ์•„์•ผ ํ•œ๋‹ค.

๋ณด์•ˆ ๊ธฐ์ˆ  ๊ธฐ์—… ํ•ด์ปค์›(HackerOne)์˜ ์„ค๋ฌธ์กฐ์‚ฌ์— ๋”ฐ๋ฅด๋ฉด, CISO์˜ 84%๊ฐ€ ํ˜„์žฌ AI ๋ณด์•ˆ์„ ๋‹ด๋‹นํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, 82%๋Š” ๋ฐ์ดํ„ฐ ํ”„๋ผ์ด๋ฒ„์‹œ๊นŒ์ง€ ๊ด€ํ• ํ•˜๊ณ  ์žˆ๋‹ค. CISO๊ฐ€ AI ์‹œ์Šคํ…œ๊ณผ ์ด๋ฅผ ํ•™์Šต์‹œํ‚ค๋Š” ๋ฐ์ดํ„ฐ๋ฅผ ๊ฒจ๋ƒฅํ•œ ๊ณต๊ฒฉ์— ๋Œ€์‘ํ•˜๋Š” ๋ณด์•ˆ ์ „๋žต์„ ๋ฐœ์ „์‹œํ‚ค์ง€ ๋ชปํ•  ๊ฒฝ์šฐ, ํ–ฅํ›„ ๋ฐœ์ƒํ•˜๋Š” ๋ฌธ์ œ๋Š” AI ์ด๋‹ˆ์…”ํ‹ฐ๋ธŒ ๊ธฐํšยท์ถœ๋ฒ” ๊ณผ์ •์— ์ฐธ์—ฌํ–ˆ๋Š”์ง€์™€ ๊ด€๊ณ„์—†์ด CISO์˜ ๋ฆฌ๋”์‹ญ์— ๋Œ€ํ•œ ์ฑ…์ž„์œผ๋กœ ์ด์–ด์งˆ ์ˆ˜ ์žˆ๋‹ค.

์ด์™€ ๊ด€๋ จํ•ด ์ปจ์Šคํ…”๋ ˆ์ด์…˜ ๋ฆฌ์„œ์น˜์˜ ๋ฉ”ํƒ€๋Š” โ€œ์„ ์ œ์ ์ธ AI ๋ณด์•ˆ ์ „๋žต์ด ํ•„์š”ํ•˜๋‹คโ€๋ผ๊ณ  ์ง„๋‹จํ–ˆ๋‹ค.

์ปจ์Šคํ…”๋ ˆ์ด์…˜ ๋ฆฌ์„œ์น˜์˜ ๋ณด๊ณ ์„œ โ€˜์ „ํ†ต์  ์‚ฌ์ด๋ฒ„ ๋ฐฉ์–ด๋ฅผ ๋„˜์–ด์„  AI ๋ณด์•ˆ: AI์™€ ์ž์œจ์„ฑ ์‹œ๋Œ€๋ฅผ ์œ„ํ•œ ์‚ฌ์ด๋ฒ„๋ณด์•ˆ ์žฌ๊ตฌ์ƒโ€™์—์„œ ์ˆ˜์„ ์• ๋„๋ฆฌ์ŠคํŠธ ์น˜๋ฝ ๋ฉ”ํƒ€๋Š” โ€œAI ๋ณด์•ˆ์€ ๋‹จ์ˆœํ•œ ๊ธฐ์ˆ ์  ๊ณผ์ œ๊ฐ€ ์•„๋‹ˆ๋ผ, ๊ฒฝ์˜์ง„์˜ ์ „ํญ์ ์ธ ์ง€์ง€์™€ ๋ถ€์„œ ๊ฐ„ ํ˜‘๋ ฅ์ด ์š”๊ตฌ๋˜๋Š” ์ „๋žต์  ๊ณผ์ œโ€๋ผ๋ฉฐ โ€œ๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค๋Š” AI ๋ณด์•ˆ์˜ ์ถœ๋ฐœ์ ์œผ๋กœ, ํ•™์Šต ๋ฐ์ดํ„ฐ์™€ ๋ชจ๋ธ ์ž…๋ ฅ์˜ ๋ฌด๊ฒฐ์„ฑ๊ณผ ์ถœ์ฒ˜๋ฅผ ๋ณด์žฅํ•˜๋Š” ๊ฒƒ์ด ํ•ต์‹ฌโ€์ด๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์ด์–ด โ€œ๋ณด์•ˆ ํŒ€์€ AI ๊ธฐ๋ฐ˜ ์œ„ํ—˜์„ ๋‹ค๋ฃจ๊ธฐ ์œ„ํ•œ ์ƒˆ๋กœ์šด ์ „๋ฌธ์„ฑ์„ ํ™•๋ณดํ•ด์•ผ ํ•˜๋ฉฐ, ๋น„์ฆˆ๋‹ˆ์Šค ๋ฆฌ๋”๋Š” ์ž์œจํ˜• AI ์‹œ์Šคํ…œ์ด ๊ฐ–๋Š” ์˜ํ–ฅ๊ณผ ์ด๋ฅผ ์ฑ…์ž„๊ฐ ์žˆ๊ฒŒ ๊ด€๋ฆฌํ•˜๊ธฐ ์œ„ํ•œ ๊ฑฐ๋ฒ„๋„Œ์Šค ์ฒด๊ณ„๋ฅผ ์ธ์‹ํ•ด์•ผ ํ•œ๋‹คโ€๋ผ๊ณ  ๋ฐํ˜”๋‹ค.

AI ์‹œ์Šคํ…œ ๊ณต๊ฒฉ์„ ํ‰๊ฐ€ยท๊ด€๋ฆฌยท๋Œ€์‘ํ•˜๊ธฐ ์œ„ํ•œ ์ „๋žต๋„ ์ ์ฐจ ๊ตฌ์ฒดํ™”๋˜๊ณ  ์žˆ๋‹ค. ๊ฐ•๋ ฅํ•œ ๋ฐ์ดํ„ฐ ๊ฑฐ๋ฒ„๋„Œ์Šค๋ฅผ ์œ ์ง€ํ•˜๋Š” ๊ฒƒ๊ณผ ํ•จ๊ป˜, ์ „๋ฌธ๊ฐ€๋“ค์€ AI ๋ชจ๋ธ์„ ๋ฐฐํฌ ์ „์— ์ ๊ฒ€ํ•˜๊ณ , ์šด์˜ ์ค‘์ธ AI ์‹œ์Šคํ…œ์„ ์ง€์†์ ์œผ๋กœ ๋ชจ๋‹ˆํ„ฐ๋งํ•˜๋ฉฐ, ๋ ˆ๋“œํŒ€์„ ํ™œ์šฉํ•ด ๋ชจ๋ธ์„ ์‹œํ—˜ํ•  ํ•„์š”๊ฐ€ ์žˆ๋‹ค๊ณ  ์กฐ์–ธํ•œ๋‹ค.

PwC์˜ ์œ„์ฝ”ํ”„๋Š” โ€œํŠน์ • ๊ณต๊ฒฉ ์œ ํ˜•์— ๋Œ€์‘ํ•˜๊ธฐ ์œ„ํ•ด CISO๊ฐ€ ๋ณ„๋„์˜ ์กฐ์น˜๋ฅผ ์ทจํ•ด์•ผ ํ•  ์ˆ˜๋„ ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค. ์˜ˆ๋ฅผ ๋“ค์–ด ๋ชจ๋ธ ํƒˆ์ทจ๋ฅผ ์‚ฌ์ „์— ์ฐจ๋‹จํ•˜๋ ค๋Š” ๊ฒฝ์šฐ, ์˜์‹ฌ์Šค๋Ÿฌ์šด ์ฟผ๋ฆฌ์™€ ํŒจํ„ด์„ ๋ชจ๋‹ˆํ„ฐ๋งํ•˜๊ณ  ์‘๋‹ต ์†๋„๋ฅผ ์ œํ•œํ•˜๊ฑฐ๋‚˜ ํƒ€์ž„์•„์›ƒ์„ ์„ค์ •ํ•˜๋Š” ๋ฐฉ์‹์ด ํ™œ์šฉ๋  ์ˆ˜ ์žˆ๋‹ค. ํšŒํ”ผ ๊ณต๊ฒฉ์„ ์ค„์ด๊ธฐ ์œ„ํ•ด์„œ๋Š” ์ ๋Œ€์  ํ•™์Šต, ์ฆ‰ ํ•ด๋‹น ๊ณต๊ฒฉ ์œ ํ˜•์— ๋Œ€๋น„ํ•˜๋„๋ก ๋ชจ๋ธ์„ ํ•™์Šต์‹œํ‚ค๋Š” ์ ‘๊ทผ๋„ ๊ณ ๋ คํ•  ์ˆ˜ ์žˆ๋‹ค.

MITRE ATLAS ๋„์ž…๋„ ํ•˜๋‚˜์˜ ๋ฐฉ์•ˆ์œผ๋กœ ์ œ์‹œ๋œ๋‹ค. ์ธ๊ณต์ง€๋Šฅ ์‹œ์Šคํ…œ์„ ๊ฒจ๋ƒฅํ•œ ์ ๋Œ€์  ์œ„ํ˜‘ ์ง€ํ˜•(Adversarial Threat Landscape for Artificial-Intelligence Systems)์˜ ์•ฝ์ž์ธ ์ด ํ”„๋ ˆ์ž„์›Œํฌ๋Š” ๊ณต๊ฒฉ์ž๊ฐ€ AI ์‹œ์Šคํ…œ์„ ์–ด๋–ป๊ฒŒ ๋…ธ๋ฆฌ๋Š”์ง€๋ฅผ ์ฒด๊ณ„์ ์œผ๋กœ ์ •๋ฆฌํ•˜๊ณ , ์ „์ˆ ยท๊ธฐ๋ฒ•ยท์ ˆ์ฐจ(TTP)๋ฅผ ์‹๋ณ„ํ•  ์ˆ˜ ์žˆ๋Š” ์ง€์‹ ๊ธฐ๋ฐ˜์„ ์ œ๊ณตํ•œ๋‹ค.

๋ณด์•ˆ ๋ฐ AI ์ „๋ฌธ๊ฐ€๋“ค์€ ์ด๋Ÿฌํ•œ ์กฐ์น˜๋ฅผ ์‹คํ–‰ํ•˜๋Š” ๋ฐ ํ˜„์‹ค์ ์ธ ์–ด๋ ค์›€์ด ๋”ฐ๋ฅธ๋‹ค๋Š” ์ ๋„ ์ธ์ •ํ•œ๋‹ค. ๋งŽ์€ CISO๊ฐ€ ์„€๋„ AI ๋ฌธ์ œ๋‚˜, ๊ณต๊ฒฉ์ž๊ฐ€ AI๋ฅผ ํ™œ์šฉํ•˜๋ฉด์„œ ๋” ๋น ๋ฅด๊ณ  ์ •๊ตํ•ด์ง„ ๊ธฐ์กด ์‚ฌ์ด๋ฒ„ ๊ณต๊ฒฉ์— ์šฐ์„  ๋Œ€์‘ํ•ด์•ผ ํ•˜๋Š” ์ƒํ™ฉ์— ๋†“์—ฌ ์žˆ๋‹ค. ์—ฌ๊ธฐ์— ๋”ํ•ด, AI ์‹œ์Šคํ…œ์„ ๊ฒจ๋ƒฅํ•œ ๊ณต๊ฒฉ์ด ์•„์ง ์ดˆ๊ธฐ ๋‹จ๊ณ„์— ๋จธ๋ฌผ๋Ÿฌ ์žˆ๊ณ  ์ผ๋ถ€๋Š” ์—ฌ์ „ํžˆ ์ด๋ก ์  ์ˆ˜์ค€์œผ๋กœ ํ‰๊ฐ€๋œ๋‹ค๋Š” ์ ๋„, ๊ด€๋ จ ์ „๋žต๊ณผ ์—ญ๋Ÿ‰์„ ํ™•๋ณดํ•˜๊ธฐ ์œ„ํ•œ ์ž์› ๋ฐฐ๋ถ„์„ ์–ด๋ ต๊ฒŒ ๋งŒ๋“œ๋Š” ์š”์ธ์ด๋‹ค.

์ƒŒ์Šค์˜ ๋ฆฌ๋Š” โ€œCISO ์ž…์žฅ์—์„œ AI ๋ฐฑ์—”๋“œ๋ฅผ ๊ฒจ๋ƒฅํ•œ ๊ณต๊ฒฉ์€ ์•„์ง ์—ฐ๊ตฌ๊ฐ€ ์ง„ํ–‰ ์ค‘์ธ ์˜์—ญ์ด์–ด์„œ ๋งค์šฐ ๊นŒ๋‹ค๋กญ๋‹คโ€๋ผ๋ฉฐ โ€œ์šฐ๋ฆฌ๋Š” ํ•ด์ปค๊ฐ€ ๋ฌด์—‡์„ ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ๊ทธ ์ด์œ ๊ฐ€ ๋ฌด์—‡์ธ์ง€๋ฅผ ํŒŒ์•…ํ•˜๋Š” ์ดˆ๊ธฐ ๋‹จ๊ณ„์— ์žˆ๋‹คโ€๋ผ๊ณ  ์„ค๋ช…ํ–ˆ๋‹ค.

๋ฆฌ์™€ ๋‹ค๋ฅธ ์ „๋ฌธ๊ฐ€๋“ค์€ ์กฐ์ง์ด AI ๊ฒฝ์Ÿ์—์„œ ๋’ค์ฒ˜์ง€์ง€ ์•Š์œผ๋ ค๋Š” ์••๋ฐ•์„ ๋ฐ›๊ณ  ์žˆ๋‹ค๋Š” ์ ์„ ์ดํ•ดํ•˜๋ฉด์„œ๋„, AI ์‹œ์Šคํ…œ ๋ณด์•ˆ์„ ์‚ฌํ›„ ๊ณผ์ œ๋กœ ๋ฐ€์–ด์„œ๋Š” ์•ˆ ๋œ๋‹ค๊ณ  ๊ฐ•์กฐํ•œ๋‹ค.

PwC ์‚ฌ์ด๋ฒ„ยท๋ฆฌ์Šคํฌ ํ˜์‹  ์—ฐ๊ตฌ์†Œ๋ฅผ ์ด๋„๋Š” ๋งท ๊ณ ํ–„์€ โ€œ์ธํ”„๋ผ๋ฅผ ๊ตฌ์ถ•ํ•˜๋Š” ๊ณผ์ •์—์„œ ์–ด๋–ค ํ˜•ํƒœ์˜ ๊ณต๊ฒฉ์ด ๊ฐ€๋Šฅํ• ์ง€๋ฅผ ํ•จ๊ป˜ ๊ณ ๋ฏผํ•˜๋Š” ๊ฒƒ์ด CISO์—๊ฒŒ ํ•ต์‹ฌ ๊ณผ์ œโ€๋ผ๊ณ  ์–ธ๊ธ‰ํ–ˆ๋‹ค.
dl-ciokorea@foundryco.com

โŒ
โŒ