Loading your language..
Anthropic Reveals Its Model Claude Is Assisting in Developing Its Next Version

Anthropic Reveals Its Model Claude Is Assisting in Developing Its Next Version

C2๐Ÿ‡ฐ๐Ÿ‡ท ํ•œ๊ตญ์–ด โ†’ ๐Ÿ‡บ๐Ÿ‡ธ English

September 18th, 2026

Anthropic Reveals Its Model Claude Is Assisting in Developing Its Next Version

C2
Please note: This article has been simplified for language learning purposes. Some context and nuance from the original text may have been modified or removed.

๐Ÿ‡บ๐Ÿ‡ธ English

In an announcement made on Thursday, this artificial intelligence laboratory declared that Anthropic's Claude is helping to support the development of an even more intelligent next-generation model.

์ด ์ธ๊ณต์ง€๋Šฅ ์—ฐ๊ตฌ์†Œ๋Š” ๋ชฉ์š”์ผ ๋ฐœํ‘œ๋ฅผ ํ†ตํ•ด, ์•คํŠธ๋กœํ”ฝ์˜ ํด๋กœ๋“œ๊ฐ€ ํ•œ์ธต ๋” ์ง€๋Šฅ์ ์ธ ์ฐจ๊ธฐ ๋ชจ๋ธ์˜ ๊ฐœ๋ฐœ์„ ์ง€์›ํ•˜๋Š” ๋ฐ ์ผ์กฐํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ๊ณตํ‘œํ•˜์˜€๋‹ค.

Claude is spearheading 26% of Anthropic's model research and development, which the company explained suggests that, under human supervision, it can complete "most tasks from start to finish from high-level prompts."

ํด๋กœ๋“œ๋Š” ์•คํŠธ๋กœํ”ฝ์˜ ๋ชจ๋ธ ์—ฐ๊ตฌ ๊ฐœ๋ฐœ ๊ฐ€์šด๋ฐ 26%๋ฅผ ์ฃผ๋„ํ•˜๊ณ  ์žˆ์œผ๋ฉฐ, ์ด๋Š” ์ธ๊ฐ„์˜ ๊ฐ๋… ์•„๋ž˜ "๊ณ ์ˆ˜์ค€ ํ”„๋กฌํ”„ํŠธ๋กœ๋ถ€ํ„ฐ ๋Œ€๋ถ€๋ถ„์˜ ์ž‘์—…์„ ์ฒ˜์Œ๋ถ€ํ„ฐ ๋๊นŒ์ง€" ์™„์ˆ˜ํ•  ์ˆ˜ ์žˆ์Œ์„ ์‹œ์‚ฌํ•œ๋‹ค๊ณ  ๋™์‚ฌ๋Š” ์„ค๋ช…ํ–ˆ๋‹ค.

Nevertheless, it is still difficult to regard the model as having reached the stage of performing tasks with complete autonomy.

๋‹ค๋งŒ ๋ชจ๋ธ์ด ์™„์ „ํ•œ ์ž์œจ์„ฑ์œผ๋กœ ์ž‘์—…์„ ์ˆ˜ํ–‰ํ•˜๋Š” ๋‹จ๊ณ„์— ์ด๋ฅด๋ €๋‹ค๊ณ  ๋ณด๊ธฐ๋Š” ์•„์ง ์–ด๋ ต๋‹ค.

Nevertheless, approximately 90% of the company's research and development is carried out through "collaboration" with Claude, which Anthropic stated suggests that the model possesses the capability to "perform large-scale tasks under meticulous human direction."

๊ทธ๋Ÿผ์—๋„ ๋ถˆ๊ตฌํ•˜๊ณ , ํšŒ์‚ฌ ์—ฐ๊ตฌ ๊ฐœ๋ฐœ์˜ ์•ฝ 90%๊ฐ€ ํด๋กœ๋“œ์™€์˜ "ํ˜‘์—…"์„ ํ†ตํ•ด ์ˆ˜ํ–‰๋˜๊ณ  ์žˆ์œผ๋ฉฐ, ์ด๋Š” ํ•ด๋‹น ๋ชจ๋ธ์ด "์ธ๊ฐ„์˜ ๋ฉด๋ฐ€ํ•œ ์ง€์‹œ ์•„๋ž˜ ๋Œ€๊ทœ๋ชจ ์ž‘์—…์„ ์ˆ˜ํ–‰"ํ•  ์ˆ˜ ์žˆ๋Š” ์—ญ๋Ÿ‰์„ ์ง€๋‹Œ๋‹ค๋Š” ๊ฒƒ์„ ์‹œ์‚ฌํ•œ๋‹ค๊ณ  ์•คํŠธ๋กœํ”ฝ์€ ๋ฐํ˜”๋‹ค.

This announcement came amid calls from a number of prominent figures in the AI industry, led by Anthropic CEO Dario Amodei, for moderating the pace of technological development under the pretext of safety concerns.

์ด ๋ฐœํ‘œ๋Š” ์•คํŠธ๋กœํ”ฝ CEO ๋‹ค๋ฆฌ์˜ค ์•„๋ชจ๋ฐ์ด๋ฅผ ํ•„๋‘๋กœ ํ•œ ์ผ๋ถ€ AI ์—…๊ณ„ ์ฃผ์š” ์ธ์‚ฌ๋“ค์ด ์•ˆ์ „ ๋ฌธ์ œ๋ฅผ ๋ช…๋ถ„์œผ๋กœ ๊ธฐ์ˆ  ๊ฐœ๋ฐœ ์†๋„ ์กฐ์ ˆ์„ ์ด‰๊ตฌํ•˜๊ณ  ์žˆ๋Š” ๊ฐ€์šด๋ฐ ๋‚˜์˜จ ๊ฒƒ์ด๋‹ค.

Given the trajectory of the pace of AI development, the company stated in a blog post that "every available means must be pursued to minimize the gap between the knowledge held by frontier labs and the knowledge accessible to the public."

AI ๊ฐœ๋ฐœ ์†๋„์˜ ์ถ”์ด๋ฅผ ๊ฐ์•ˆํ•  ๋•Œ, "์ตœ์ „์„  ์—ฐ๊ตฌ์†Œ๊ฐ€ ๋ณด์œ ํ•œ ์ง€์‹๊ณผ ๋Œ€์ค‘์ด ์ ‘๊ทผ ๊ฐ€๋Šฅํ•œ ์ง€์‹ ๊ฐ„์˜ ๊ฒฉ์ฐจ๋ฅผ ์ตœ์†Œํ™”ํ•˜๊ธฐ ์œ„ํ•ด ๊ฐ€์šฉํ•œ ๋ชจ๋“  ์ˆ˜๋‹จ์„ ๊ฐ•๊ตฌํ•ด์•ผ ํ•œ๋‹ค"๊ณ  ํšŒ์‚ฌ๋Š” ๋ธ”๋กœ๊ทธ ๊ฒŒ์‹œ๋ฌผ์„ ํ†ตํ•ด ๋ฐํ˜”๋‹ค.

"This entails measuring AI development with greater precision, disclosing it transparently, and affording society the opportunity to determine how this information should be used."

"์ด๋Š” AI ๊ฐœ๋ฐœ์„ ๋ณด๋‹ค ์ •๋ฐ€ํ•˜๊ฒŒ ์ธก์ •ํ•˜๊ณ , ์ด๋ฅผ ํˆฌ๋ช…ํ•˜๊ฒŒ ๊ณต๊ฐœํ•˜๋ฉฐ, ์‚ฌํšŒ๊ฐ€ ์ด ์ •๋ณด๋ฅผ ์–ด๋–ป๊ฒŒ ํ™œ์šฉํ• ์ง€ ๊ฒฐ์ •ํ•  ์ˆ˜ ์žˆ๋Š” ๊ธฐํšŒ๋ฅผ ๋ถ€์—ฌํ•˜๋Š” ๊ฒƒ์„ ์˜๋ฏธํ•œ๋‹ค."

The company stated that if models were to accelerate their own development, "it could become increasingly difficult for humans to understand or control such systems."

ํšŒ์‚ฌ๋Š” ๋ชจ๋ธ์ด ์Šค์Šค๋กœ์˜ ๊ฐœ๋ฐœ์„ ๊ฐ€์†ํ™”ํ•  ๊ฒฝ์šฐ "์ธ๊ฐ„์ด ์ด๋Ÿฌํ•œ ์‹œ์Šคํ…œ์„ ์ดํ•ดํ•˜๊ฑฐ๋‚˜ ํ†ต์ œํ•˜๊ธฐ ๋” ์–ด๋ ค์›Œ์งˆ ์ˆ˜ ์žˆ๋‹ค"๊ณ  ๋ฐํ˜”๋‹ค.

It contended that sharing such metrics could contribute to a more precise understanding of how close leading AI laboratories have come to recursive self-improvementโ€”that is, the capacity of models to autonomously construct their successors.

์ด๋Ÿฌํ•œ ์ง€ํ‘œ๋ฅผ ๊ณต์œ ํ•˜๋Š” ๊ฒƒ์€ ์ฃผ์š” AI ์—ฐ๊ตฌ์†Œ๋“ค์ด ์žฌ๊ท€์  ์ž๊ธฐ ๊ฐœ์„ , ์ฆ‰ ๋ชจ๋ธ์ด ์ž์œจ์ ์œผ๋กœ ํ›„์† ๋ชจ๋ธ์„ ๊ตฌ์ถ•ํ•˜๋Š” ๋Šฅ๋ ฅ์— ์–ผ๋งˆ๋‚˜ ๊ทผ์ ‘ํ–ˆ๋Š”์ง€๋ฅผ ๋ณด๋‹ค ์ •ํ™•ํžˆ ํŒŒ์•…ํ•˜๋Š” ๋ฐ ๊ธฐ์—ฌํ•  ์ˆ˜ ์žˆ๋‹ค๊ณ  ์ฃผ์žฅํ–ˆ๋‹ค.

Anthropic further urged other AI developers to share similar metrics on a regular basis, while encouraging the adoption of open methodologies so that, over time, comparisons of figures across laboratories could potentially become feasible.

์•คํŠธ๋กœํ”ฝ์€ ๋‚˜์•„๊ฐ€ ์—ฌํƒ€ AI ๊ฐœ๋ฐœ์ž๋“ค์—๊ฒŒ๋„ ์œ ์‚ฌํ•œ ์ง€ํ‘œ๋ฅผ ์ •๊ธฐ์ ์œผ๋กœ ๊ณต์œ ํ•  ๊ฒƒ์„ ์ด‰๊ตฌํ•˜๋Š” ํ•œํŽธ, ์‹œ๊ฐ„์ด ๊ฒฝ๊ณผํ•จ์— ๋”ฐ๋ผ ์ž ์žฌ์ ์œผ๋กœ ์—ฐ๊ตฌ์†Œ ๊ฐ„ ์ˆ˜์น˜ ๋น„๊ต๊ฐ€ ๊ฐ€๋Šฅํ•ด์งˆ ์ˆ˜ ์žˆ๋„๋ก ๊ณต๊ฐœ ๋ฐฉ๋ฒ•๋ก ์˜ ์ฑ„ํƒ์„ ์žฅ๋ คํ–ˆ๋‹ค.

While Anthropic's publicly available materials alone offered no means of gauging how close the company had come to achieving recursive self-improvement, the pace at which Claude was contributing incrementally to research and development was nonetheless noteworthy.

์•คํŠธ๋กœํ”ฝ์˜ ๊ณต๊ฐœ ์ž๋ฃŒ๋งŒ์œผ๋กœ๋Š” ํ•ด๋‹น ๊ธฐ์—…์ด ์žฌ๊ท€์  ์ž๊ธฐ ๊ฐœ์„ ์˜ ๋‹ฌ์„ฑ์— ์–ผ๋งˆ๋‚˜ ๊ทผ์ ‘ํ–ˆ๋Š”์ง€ ๊ฐ€๋Š ํ•  ์ˆ˜ ์—†์—ˆ์œผ๋‚˜, ํด๋กœ๋“œ๊ฐ€ ์—ฐ๊ตฌ ๊ฐœ๋ฐœ์— ์ ์ฆ์ ์œผ๋กœ ๊ธฐ์—ฌํ•˜๋Š” ์†๋„๋งŒํผ์€ ์ฃผ๋ชฉํ•  ๋งŒํ•˜๋‹ค.

The proportion of tasks that Claude "led" โ€” that is, carried out for the most part under human supervision โ€” was nil in February.

ํด๋กœ๋“œ๊ฐ€ "์ฃผ๋„"ํ•˜๋Š”, ์ฆ‰ ์ธ๊ฐ„์˜ ๊ฐ๋… ํ•˜์— ๊ทธ ๋Œ€๋ถ€๋ถ„์„ ์ˆ˜ํ–‰ํ•˜๋Š” ์ž‘์—…์˜ ๋น„์œจ์€ 2์›”์—๋Š” ์ „๋ฌดํ–ˆ๋‹ค.

By August, six months later, it had reached this threshold, leading a quarter of research and development tasks.

6๊ฐœ์›” ํ›„์ธ 8์›”์— ์ด๋ฅด๋Ÿฌ์„œ๋Š” ์—ฐ๊ตฌ ๊ฐœ๋ฐœ ์ž‘์—…์˜ 4๋ถ„์˜ 1์„ ์ฃผ๋„ํ•จ์œผ๋กœ์จ ์ด ๊ธฐ์ค€์— ๋„๋‹ฌํ–ˆ๋‹ค.

The company also disclosed granular details concerning the agent oversight measures currently in force, stating that, as of August, approximately 30,000 agents were engaged in research and engineering tasks.

ํšŒ์‚ฌ๋Š” ๋˜ํ•œ ์‹œํ–‰ ์ค‘์ธ ์—์ด์ „ํŠธ ๊ฐ๋… ์กฐ์น˜์— ๊ด€ํ•œ ์„ธ๋ถ€ ์ •๋ณด๋ฅผ ๊ณต๊ฐœํ•˜๋ฉฐ, 8์›” ๊ธฐ์ค€ ์•ฝ 30,000๊ฐœ์˜ ์—์ด์ „ํŠธ๊ฐ€ ์—ฐ๊ตฌ ๋ฐ ์—”์ง€๋‹ˆ์–ด๋ง ์ž‘์—…์„ ์ˆ˜ํ–‰ํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค.

The company noted that these oversight measures are of pivotal significance in ascertaining how frequently the monitoring system detects errant agent behaviour.

๊ฐ๋… ์กฐ์น˜๋Š” ๋ชจ๋‹ˆํ„ฐ๋ง ์‹œ์Šคํ…œ์ด ์—์ด์ „ํŠธ์˜ ์ž˜๋ชป๋œ ํ–‰๋™์„ ์–ผ๋งˆ๋‚˜ ๋นˆ๋ฒˆํžˆ ๊ฐ์ง€ํ•˜๋Š”์ง€ ํŒŒ์•…ํ•˜๋Š” ๋ฐ ์ค‘๋Œ€ํ•œ ์˜์˜๋ฅผ ์ง€๋‹Œ๋‹ค๊ณ  ํšŒ์‚ฌ๋Š” ์–ธ๊ธ‰ํ–ˆ๋‹ค.

Anthropic has recently committed to embedding external third-party evaluators in-house to scrutinise its safety endeavours.

์•คํŠธ๋กœํ”ฝ์€ ์ตœ๊ทผ ์ž์‚ฌ์˜ ์•ˆ์ „ ๋…ธ๋ ฅ์„ ๊ฐ์‹œํ•  ์™ธ๋ถ€ ์ œ3์ž ํ‰๊ฐ€์ž๋ฅผ ์‚ฌ๋‚ด์— ๋ฐฐ์น˜ํ•˜๊ฒ ๋‹ค๊ณ  ๊ณต์•ฝํ–ˆ๋‹ค.

Since a researcher affiliated with Anthropic resigned last week after issuing a stern warning about the threats that technology poses to humanity, discussions surrounding AI safety have recently been ignited on a broad scale.

์ง€๋‚œ์ฃผ ์•คํŠธ๋กœํ”ฝ ์†Œ์† ์—ฐ๊ตฌ์› ํ•œ ๋ช…์ด ๊ธฐ์ˆ ์ด ์ธ๋ฅ˜์— ์ œ๊ธฐํ•˜๋Š” ์œ„ํ˜‘์— ๊ด€ํ•œ ์ค€์—„ํ•œ ๊ฒฝ๊ณ ๋ฅผ ๋‚จ๊ธฐ๊ณ  ์‚ฌ์ž„ํ•œ ์ด๋ž˜, AI ์•ˆ์ „์„ ๋‘˜๋Ÿฌ์‹ผ ๋…ผ์˜๊ฐ€ ์ตœ๊ทผ ๊ด‘๋ฒ”์œ„ํ•˜๊ฒŒ ์ด‰๋ฐœ๋˜์—ˆ๋‹ค.

Subsequently, Amodei, OpenAI CEO Sam Altman, Elon Musk, and other technology leaders voiced their support for the notion of moderating the pace of development, whereas other technology leaders and President Donald Trump maintained an opposing stance.

๊ทธ ํ›„ ์•„๋ชจ๋ฐ์ด, ์ƒ˜ ์˜ฌํŠธ๋จผ OpenAI CEO, ์ผ๋ก  ๋จธ์Šคํฌ๋ฅผ ๋น„๋กฏํ•œ ์—ฌํƒ€ ๊ธฐ์ˆ  ๋ฆฌ๋”๋“ค์€ ๊ฐœ๋ฐœ ์†๋„ ์กฐ์ ˆ์ด๋ผ๋Š” ๊ตฌ์ƒ์— ์ง€์ง€๋ฅผ ํ‘œ๋ช…ํ•œ ๋ฐ˜๋ฉด, ๋‹ค๋ฅธ ๊ธฐ์ˆ  ๋ฆฌ๋”๋“ค๊ณผ ๋„๋„๋“œ ํŠธ๋Ÿผํ”„ ๋Œ€ํ†ต๋ น์€ ์ด์— ๋ฐ˜๋Œ€ํ•˜๋Š” ์ž…์žฅ์„ ๊ฒฌ์ง€ํ–ˆ๋‹ค.

September 18th, 2026

Trending Articles

Warren Buffett steps down as Berkshire Hathaway chairman in the next phase of succession planning

Warren Buffett steps down as Berkshire Hathaway chairman in the next phase of succession planning

์›Œ๋Ÿฐ ๋ฒ„ํ•, ํ›„๊ณ„ ๊ณ„ํš์˜ ๋‹ค์Œ ๋‹จ๊ณ„๋กœ ๋ฒ„ํฌ์…” ํ•ด์„œ์›จ์ด ํšŒ์žฅ์ง ์‚ฌ์ž„

C2Sep 18
Trump signals White House press ban for CNN, MS NOW, and Politicoโ€ฆ widening targeting of news outlets

Trump signals White House press ban for CNN, MS NOW, and Politicoโ€ฆ widening targeting of news outlets

ํŠธ๋Ÿผํ”„, CNNยทMS NOWยทํด๋ฆฌํ‹ฐ์ฝ” ๋ฐฑ์•…๊ด€ ์ถœ์ž… ๊ธˆ์ง€ ์˜ˆ๊ณ โ€ฆ์–ธ๋ก  ๋งค์ฒด ๊ฒจ๋ƒฅ ํ™•๋Œ€

C2Sep 18
Bank of Japan raises benchmark interest rate to 1.25% โ€” highest in 31 years

Bank of Japan raises benchmark interest rate to 1.25% โ€” highest in 31 years

์ผ๋ณธ ์ค‘์•™์€ํ–‰, ๊ธฐ์ค€๊ธˆ๋ฆฌ 1.25%๋กœ ์ธ์ƒ โ€” 31๋…„ ๋งŒ์˜ ์ตœ๊ณ ์น˜

C2Sep 18
The Federal Reserve's Interest Rate Hike: Background and Implications

The Federal Reserve's Interest Rate Hike: Background and Implications

์—ฐ์ค€์˜ ๊ธˆ๋ฆฌ ์ธ์ƒ: ๋ฐฐ๊ฒฝ๊ณผ ํ•จ์˜

C2Sep 17
Fed, in first rate hike in three years, directly counters Trump's demand for a cut

Fed, in first rate hike in three years, directly counters Trump's demand for a cut

์—ฐ์ค€, 3๋…„ ๋งŒ์˜ ๊ธฐ์ค€๊ธˆ๋ฆฌ ์ธ์ƒ์œผ๋กœ ํŠธ๋Ÿผํ”„์˜ ์ธํ•˜ ์š”๊ตฌ์— ์ •๋ฉด ๋Œ€์‘

C2Sep 17
Jobless Claims Fall to 196,000, Lowest Since Mid-July...Layoffs Remain Subdued

Jobless Claims Fall to 196,000, Lowest Since Mid-July...Layoffs Remain Subdued

Jobless Claims Fall to 196,000, Lowest Since Mid-July...Layoffs Remain Subdued

C2Sep 17
Trump, characterizing AI risk as a 'hoax,' claims the opposition to AI and data centers is a 'pathological conspiracy'

Trump, characterizing AI risk as a 'hoax,' claims the opposition to AI and data centers is a 'pathological conspiracy'

ํŠธ๋Ÿผํ”„, AI ์œ„ํ—˜์„ '์‚ฌ๊ธฐ'๋กœ ๊ทœ์ •ํ•˜๋ฉฐ AIยท๋ฐ์ดํ„ฐ ์„ผํ„ฐ ๋ฐ˜๋Œ€ํ•˜๋Š” '๋ณ‘์ ์ธ ์Œ๋ชจ' ์ฃผ์žฅ

C2Sep 14
Supreme Court to Decide Whether to Hear Maine Lobstermen's Challenge to GPS Boat-Tracking Mandate

Supreme Court to Decide Whether to Hear Maine Lobstermen's Challenge to GPS Boat-Tracking Mandate

Supreme Court to Decide Whether to Hear Maine Lobstermen's Challenge to GPS Boat-Tracking Mandate

C2Sep 14
Houthi rebels seize two strategic Red Sea islands... An overview of the Middle East situation

Houthi rebels seize two strategic Red Sea islands... An overview of the Middle East situation

ํ›„ํ‹ฐ ๋ฐ˜๊ตฐ, ์ „๋žต์  ํ™ํ•ด ๋„์„œ 2๊ณณ ์žฅ์•…... ์ค‘๋™ ์ •์„ธ ๊ฐœ๊ด€

C2Sep 14