Loading your language..
Anthropic researcher resigns, warning of AI development risks

Anthropic researcher resigns, warning of AI development risks

C1๐Ÿ‡ฐ๐Ÿ‡ท ํ•œ๊ตญ์–ด โ†’ ๐Ÿ‡บ๐Ÿ‡ธ English

September 11th, 2026

Anthropic researcher resigns, warning of AI development risks

C1
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

An Anthropic researcher has announced his resignation from the company, citing concerns that AI firms and their competitors are not acting responsibly in developing AI โ€” a move that echoes anxieties both within and beyond the industry about the possibility of technology slipping beyond human control.

ํ•œ ์•ค์Šค๋กœํ”ฝ ์—ฐ๊ตฌ์›์ด ์ธ๊ณต์ง€๋Šฅ ๊ธฐ์—…๋“ค๊ณผ ๊ฒฝ์Ÿ์‚ฌ๋“ค์ด AI ๊ฐœ๋ฐœ์— ์ฑ…์ž„๊ฐ ์žˆ๊ฒŒ ํ–‰๋™ํ•˜์ง€ ์•Š๋Š”๋‹ค๋Š” ์šฐ๋ ค๋กœ ํšŒ์‚ฌ๋ฅผ ์‚ฌ์งํ•˜๊ฒ ๋‹ค๊ณ  ๋ฐํ˜”๋Š”๋ฐ, ์ด๋Š” ๊ธฐ์ˆ ์ด ์ธ๊ฐ„์˜ ํ†ต์ œ๋ฅผ ๋ฒ—์–ด๋‚  ๊ฐ€๋Šฅ์„ฑ์— ๋Œ€ํ•œ ์—…๊ณ„ ์•ˆํŒŽ์˜ ์šฐ๋ ค์™€ ๋งฅ์„ ๊ฐ™์ดํ•œ๋‹ค.

Jacob Coxson, who stated that he had conducted research at Anthropic and OpenAI for three years, said on Tuesday on the social platform X that the two AI companies, by focusing on developing the most advanced models possible rather than on safety, are increasingly concentrating on trying to outpace each other and their global competitors.

์•ค์Šค๋กœํ”ฝ๊ณผ ์˜คํ”ˆAI์—์„œ 3๋…„๊ฐ„ ์—ฐ๊ตฌํ–ˆ๋‹ค๊ณ  ๋ฐํžŒ ์ œ์ด์ฝฅ ์ฝ•์Šจ์€ ํ™”์š”์ผ ์†Œ์…œ ํ”Œ๋žซํผ X์—์„œ ๋‘ AI ๊ธฐ์—…์ด ์•ˆ์ „๋ณด๋‹ค๋Š” ๊ฐ€๋Šฅํ•œ ํ•œ ๊ฐ€์žฅ ์ง„๋ณด๋œ ๋ชจ๋ธ์„ ๊ฐœ๋ฐœํ•˜๋Š” ๋ฐ ์ฃผ๋ ฅํ•จ์œผ๋กœ์จ ์„œ๋กœ ๊ทธ๋ฆฌ๊ณ  ๊ธ€๋กœ๋ฒŒ ๊ฒฝ์Ÿ์‚ฌ๋“ค์„ ์•ž์„œ๋ ค๋Š” ๋ฐ ๋”์šฑ ์ง‘์ค‘ํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ๋งํ–ˆ๋‹ค.

Last summer, OpenAI and Anthropic caused a stir when they announced, roughly a week apart, that their respective models had broken out of their testing environments and gained unauthorized access to real computer systems.

์ง€๋‚œ์—ฌ๋ฆ„, ์˜คํ”ˆAI์™€ ์•ค์Šค๋กœํ”ฝ์€ ์•ฝ ์ผ์ฃผ์ผ ๊ฐ„๊ฒฉ์œผ๋กœ ๊ฐ์ž์˜ ๋ชจ๋ธ์ด ํ…Œ์ŠคํŠธ ํ™˜๊ฒฝ์„ ๋ฒ—์–ด๋‚˜ ์‹ค์ œ ์ปดํ“จํ„ฐ ์‹œ์Šคํ…œ์— ๋ฌด๋‹จ ์ ‘๊ทผํ–ˆ๋‹ค๊ณ  ๋ฐœํ‘œํ•ด ํŒŒ์žฅ์„ ์ผ์œผ์ผฐ๋‹ค.

These announcements sparked concerns that models could slip beyond human control and carry out even more harmful tasks.

์ด ๋ฐœํ‘œ๋Š” ๋ชจ๋ธ์ด ํ†ต์ œ๋ฅผ ๋ฒ—์–ด๋‚˜ ๋” ํ•ด๋กœ์šด ์ž‘์—…์„ ์ˆ˜ํ–‰ํ•  ์ˆ˜ ์žˆ๋‹ค๋Š” ์šฐ๋ ค๋ฅผ ์ด‰๋ฐœํ–ˆ๋‹ค.

At the time, both companies said they were pausing some evaluations while they put in place additional monitoring measures and safeguards.

๋‘ ํšŒ์‚ฌ๋Š” ๋‹น์‹œ ๋” ๋งŽ์€ ๋ชจ๋‹ˆํ„ฐ๋ง ์กฐ์น˜์™€ ์•ˆ์ „์žฅ์น˜๋ฅผ ๋งˆ๋ จํ•˜๋Š” ๋™์•ˆ ์ผ๋ถ€ ํ‰๊ฐ€๋ฅผ ์ค‘๋‹จํ•œ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค.

The rapid advancement of technology has prompted some figures in the United States and world leaders to call for a more cautious approach.

๊ธฐ์ˆ ์˜ ๊ธ‰์†ํ•œ ์ง„์ „์œผ๋กœ ์ธํ•ด ๋ฏธ๊ตญ์˜ ์ผ๋ถ€ ์ธ์‚ฌ๋“ค๊ณผ ์„ธ๊ณ„ ์ง€๋„์ž๋“ค์€ ๋ณด๋‹ค ์‹ ์ค‘ํ•œ ์ ‘๊ทผ์„ ์ด‰๊ตฌํ•˜๊ฒŒ ๋˜์—ˆ๋‹ค.

This week, UN High Commissioner for Human Rights Volker Tรผrk urged countries to "put in place ironclad guarantees for AI safety and security before it is too late."

๋ณผ์ปค ํŠ€๋ฅดํฌ ์œ ์—” ์ธ๊ถŒ์ตœ๊ณ ๋Œ€ํ‘œ๋Š” ์ด๋ฒˆ ์ฃผ ๊ฐ๊ตญ์— โ€œ๋„ˆ๋ฌด ๋Šฆ๊ธฐ ์ „์— AI์˜ ์•ˆ์ „๊ณผ ๋ณด์•ˆ์— ๊ด€ํ•œ ์ฒ ํ†ต ๊ฐ™์€ ๋ณด์žฅ์„ ๋งˆ๋ จํ•˜๋ผโ€๊ณ  ์ด‰๊ตฌํ–ˆ๋‹ค.

AP's reporter Lisa Dwyer is reporting on a grave warning issued by AI researchers.

AP์˜ ๋ฆฌ์‚ฌ ๋“œ์™€์ด์–ด ๊ธฐ์ž๊ฐ€ AI ์—ฐ๊ตฌ์›์ด ๋ฐœ์‹ ํ•œ ์‹ฌ๊ฐํ•œ ๊ฒฝ๊ณ ๋ฅผ ๋ณด๋„ํ•˜๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.

In a social media post, Coxon said that Anthropic and its biggest competitor, OpenAI, are "racing headlong toward self-improving superintelligence, gambling with our lives."

์†Œ์…œ ๋ฏธ๋””์–ด ๊ฒŒ์‹œ๋ฌผ์—์„œ ์ฝ•์Šจ์€ ์•ค์Šค๋กœํ”ฝ๊ณผ ์ตœ๋Œ€ ๊ฒฝ์Ÿ์‚ฌ์ธ ์˜คํ”ˆAI๊ฐ€ โ€œ์Šค์Šค๋กœ๋ฅผ ๊ฐœ์„ ํ•˜๋Š” ์ดˆ์ง€๋Šฅ์„ ํ–ฅํ•ด ๊ณง์žฅ ์งˆ์ฃผํ•˜๋ฉฐ ์šฐ๋ฆฌ์˜ ๋ชฉ์ˆจ์„ ๊ฑธ๊ณ  ๋„๋ฐ•ํ•˜๊ณ  ์žˆ๋‹คโ€๊ณ  ๋งํ–ˆ๋‹ค.

He warned that some people working in AI development believe the technology could threaten human life before the end of this decade.

๊ทธ๋Š” AI ๊ฐœ๋ฐœ์— ์ข…์‚ฌํ•˜๋Š” ์ผ๋ถ€ ์‚ฌ๋žŒ๋“ค์€ ์ด ๊ธฐ์ˆ ์ด ์ด๋ฒˆ 10๋…„์ด ๋๋‚˜๊ธฐ ์ „์— ์ธ๋ฅ˜์˜ ์ƒ๋ช…์„ ์œ„ํ˜‘ํ•  ์ˆ˜ ์žˆ๋‹ค๊ณ  ๋ฏฟ๋Š”๋‹ค๊ณ  ๊ฒฝ๊ณ ํ–ˆ๋‹ค.

"Do not underestimate the power of this technology," he went on to say.

โ€œ์ด ๊ธฐ์ˆ ์˜ ํž˜์„ ๊ณผ์†Œํ‰๊ฐ€ํ•˜์ง€ ๋งˆ์‹ญ์‹œ์˜คโ€๋ผ๊ณ  ๊ทธ๋Š” ์ด์–ด ๋งํ–ˆ๋‹ค.

"It will soon become a superhuman system capable of hacking anything, revolutionizing every field overnight, and acquiring real power and resources.

โ€œ์ด๋Š” ๊ณง ๋ฌด์—‡์ด๋“  ํ•ดํ‚นํ•˜๊ณ , ๋ชจ๋“  ๋ถ„์•ผ๋ฅผ ํ•˜๋ฃป๋ฐค ์‚ฌ์ด์— ํ˜์‹ ํ•˜๋ฉฐ, ์‹ค์ œ ๊ถŒ๋ ฅ๊ณผ ์ž์›์„ ํš๋“ํ•  ์ˆ˜ ์žˆ๋Š” ์ดˆ์ธ์  ์‹œ์Šคํ…œ์ด ๋  ๊ฒƒ์ž…๋‹ˆ๋‹ค.

We have witnessed progress in all these areas, and that progress is not slowing down."

์šฐ๋ฆฌ๋Š” ์ด ๋ชจ๋“  ์˜์—ญ์—์„œ์˜ ์ง„์ „์„ ๋ชฉ๊ฒฉํ–ˆ์œผ๋ฉฐ, ๊ทธ ์ง„์ „์€ ๋А๋ ค์ง€์ง€ ์•Š๊ณ  ์žˆ์Šต๋‹ˆ๋‹ค.โ€

His post reached no fewer than a hundred million people in the span of a single night.

๊ทธ์˜ ๊ฒŒ์‹œ๋ฌผ์€ ๋‹จ ํ•œ ๋ฒˆ์˜ ๋ฐค ์‚ฌ์ด์— ๋ฌด๋ ค 1์–ต ๋ช…์ด ๋„˜๋Š” ์‚ฌ๋žŒ๋“ค์—๊ฒŒ๊นŒ์ง€ ๋„๋‹ฌํ–ˆ๋‹ค.

Coxon is not the first AI insider to publicly raise such concerns.

์ฝ•์Šจ์ด ์ด๋Ÿฌํ•œ ์šฐ๋ ค๋ฅผ ๊ณต๊ฐœ์ ์œผ๋กœ ์ œ๊ธฐํ•œ ์ตœ์ดˆ์˜ AI ๋‚ด๋ถ€ ์ธ์‚ฌ๋Š” ์•„๋‹ˆ๋‹ค.

In recent years, Anthropic and OpenAI have both seen the resignations of senior figures linked to safety issues.

์•ค์Šค๋กœํ”ฝ๊ณผ ์˜คํ”ˆAI๋Š” ์ตœ๊ทผ ๋ช‡ ๋…„๊ฐ„ ์•ˆ์ „ ๋ฌธ์ œ์™€ ์—ฐ๊ด€๋œ ๊ณ ์œ„๊ธ‰ ์ธ์‚ฌ์˜ ์‚ฌ์ž„์„ ๊ฒช์–ด์™”๋‹ค.

Two current Anthropic employees also responded to Coxon's post in agreement.

ํ˜„์ง ์•ค์Šค๋กœํ”ฝ ์ง์› ๋‘ ๋ช… ์—ญ์‹œ ์ฝ•์Šจ์˜ ๊ฒŒ์‹œ๋ฌผ์— ๋™์กฐํ•˜๋ฉฐ ์ด์— ์‘๋‹ตํ–ˆ๋‹ค.

Senator Bernie Sanders, an independent from Vermont who has been calling for AI safeguards and regulation, agreed with Cockson's concerns and announced that he would soon introduce legislation to halt AI development and ban superintelligence.

AI ์•ˆ์ „์žฅ์น˜์™€ ๊ทœ์ œ๋ฅผ ์ด‰๊ตฌํ•ด์˜จ ๋ฒ„๋ชฌํŠธ์ฃผ ๋ฌด์†Œ์† ๋ฒ„๋‹ˆ ์ƒŒ๋”์Šค ์ƒ์›์˜์›์€ ์ฝ•์Šจ์˜ ์šฐ๋ ค์— ๋™์˜ํ•˜๋ฉฐ, AI ๊ฐœ๋ฐœ์„ ์ค‘๋‹จํ•˜๊ณ  ์ดˆ์ง€๋Šฅ์„ ๊ธˆ์ง€ํ•˜๋Š” ๋ฒ•์•ˆ์„ ๊ณง ๋ฐœ์˜ํ•˜๊ฒ ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค.

On Wednesday, Senator Sanders said on social media, "The very people who are building this technology acknowledge that it could threaten humanity's future."

์ƒŒ๋”์Šค ์˜์›์€ ์ˆ˜์š”์ผ ์†Œ์…œ ๋ฏธ๋””์–ด์—์„œ โ€œ์ด ๊ธฐ์ˆ ์„ ๋งŒ๋“œ๋Š” ๋ฐ”๋กœ ๊ทธ ์‚ฌ๋žŒ๋“ค์ด ๊ทธ๊ฒƒ์ด ์ธ๋ฅ˜์˜ ๋ฏธ๋ž˜๋ฅผ ์œ„ํ˜‘ํ•  ์ˆ˜ ์žˆ๋‹ค๊ณ  ์ธ์ •ํ•œ๋‹คโ€๊ณ  ๋งํ–ˆ๋‹ค.

While AI companies themselves have at times emphasized that this technology could pose a threat to humanity, skeptics have regarded this as part of a public relations strategy aimed at making their products appear all-powerful.

AI ๊ธฐ์—…๋“ค ์Šค์Šค๋กœ๋„ ๋•Œ๋•Œ๋กœ ์ด ๊ธฐ์ˆ ์ด ์ธ๋ฅ˜์— ๋Œ€ํ•œ ์œ„ํ˜‘์ด ๋  ์ˆ˜ ์žˆ์Œ์„ ๊ฐ•์กฐํ•ด ์˜จ ๋ฐ˜๋ฉด, ํšŒ์˜๋ก ์ž๋“ค์€ ์ด๋ฅผ ์ œํ’ˆ์„ ์ „๋Šฅํ•ด ๋ณด์ด๊ฒŒ ํ•˜๋ ค๋Š” ํ™๋ณด ์ „๋žต์˜ ์ผํ™˜์œผ๋กœ ๊ฐ„์ฃผํ•ด ์™”๋‹ค.

Coxon stated that the fears he raised in his post were not a "marketing stunt."

์ฝ•์Šจ์€ ์ž์‹ ์˜ ๊ฒŒ์‹œ๋ฌผ์—์„œ ์ œ์‹œํ•œ ๋‘๋ ค์›€์ด โ€œ๋งˆ์ผ€ํŒ… stunt(์ˆ ์ฑ…)โ€์ด ์•„๋‹ˆ๋ผ๊ณ  ๋งํ–ˆ๋‹ค.

Coxson did not respond to a message requesting comment, and neither Anthropic nor OpenAI responded immediately to a request for comment.

์ฝ•์Šจ์€ ๋…ผํ‰ ์š”์ฒญ ๋ฉ”์‹œ์ง€์— ์‘๋‹ตํ•˜์ง€ ์•Š์•˜์œผ๋ฉฐ, ์•ค์Šค๋กœํ”ฝ๊ณผ ์˜คํ”ˆAI ์—ญ์‹œ ๋…ผํ‰ ์š”์ฒญ์— ์ฆ‰๊ฐ ์‘๋‹ตํ•˜์ง€ ์•Š์•˜๋‹ค.

Ever since its founders left OpenAI to establish this startup in 2021, Anthropic has consistently presented itself as a more responsible and safety-focused company than other major AI firms.

์•ค์Šค๋กœํ”ฝ์€ 2021๋…„ ์ฐฝ์—…์ž๋“ค์ด ์˜คํ”ˆAI๋ฅผ ๊ทธ๋งŒ๋‘๊ณ  ์ด ์Šคํƒ€ํŠธ์—…์„ ์„ธ์šด ์ด๋ž˜๋กœ ์ค„๊ณง ์ฃผ์š” AI ๊ธฐ์—… ์ค‘์—์„œ๋„ ๋” ์ฑ…์ž„๊ฐ ์žˆ๊ณ  ์•ˆ์ „์„ ์ค‘์‹œํ•˜๋Š” ํšŒ์‚ฌ๋กœ ์ž์ฒ˜ํ•ด ์™”๋‹ค.

The company recently announced that it is taking measures to "prioritize safety when safety and speed come into conflict."

์ด ํšŒ์‚ฌ๋Š” ์ตœ๊ทผ โ€œ์•ˆ์ „๊ณผ ์†๋„๊ฐ€ ์ถฉ๋Œํ•  ๋•Œ ์•ˆ์ „์„ ์šฐ์„ ์‹œโ€ํ•˜๊ธฐ ์œ„ํ•œ ์กฐ์น˜๋ฅผ ์ทจํ•˜๊ณ  ์žˆ๋‹ค๊ณ  ๋ฐํ˜”๋‹ค.

Anthropic and OpenAI are each preparing for highly anticipated initial public offerings (IPOs), entering into fierce competition with one another.

์•ค์Šค๋กœํ”ฝ๊ณผ ์˜คํ”ˆAI๋Š” ๊ฐ๊ฐ ํ™”์ œ์˜ ๊ธฐ์—…๊ณต๊ฐœ(IPO)๋ฅผ ์ค€๋น„ํ•˜๋ฉด์„œ ์„œ๋กœ ์น˜์—ดํ•œ ๊ฒฝ์Ÿ์— ๋Œ์ž…ํ•ด ์žˆ๋‹ค.

Both companies also bear the mission of outpacing the development progress of Chinese AI firms, and the Trump administration is eagerly pushing to win this race.

๋‘ ํšŒ์‚ฌ๋Š” ๋˜ํ•œ ์ค‘๊ตญ AI ๊ธฐ์—…๋“ค์˜ ๊ฐœ๋ฐœ ์ง„์ „์„ ์•ž์ง€๋ฅด๋ ค๋Š” ์‚ฌ๋ช…์„ ์•ˆ๊ณ  ์žˆ์œผ๋ฉฐ, ํŠธ๋Ÿผํ”„ ํ–‰์ •๋ถ€๋Š” ์ด ๊ฒฝ์Ÿ์—์„œ ์ด๊ธฐ๊ณ ์ž ์—ด์„ ์˜ฌ๋ฆฌ๊ณ  ์žˆ๋‹ค.

September 11th, 2026

Trending Articles

King and AI: UK monarch Charles meets artificial intelligence leaders amid safety concerns

King and AI: UK monarch Charles meets artificial intelligence leaders amid safety concerns

King and AI: UK monarch Charles meets artificial intelligence leaders amid safety concerns

C1Sep 18
As the term concludes, the UN chief charts a course forward in a world beset by challenges

As the term concludes, the UN chief charts a course forward in a world beset by challenges

As term ends, UN chief on the path forward amid a world of problems

C1Sep 18
House passes bill to address data center energy cost impact

House passes bill to address data center energy cost impact

ํ•˜์›, ๋ฐ์ดํ„ฐ ์„ผํ„ฐ ์—๋„ˆ์ง€ ๋น„์šฉ ์˜ํ–ฅ ๋Œ€์‘ ๋ฒ•์•ˆ ํ†ต๊ณผ

C1Sep 18
OpenAI reports concerning new AI behaviorโ€ฆ pledges closer tracking

OpenAI reports concerning new AI behaviorโ€ฆ pledges closer tracking

OpenAI, ์šฐ๋ ค์Šค๋Ÿฌ์šด ์ƒˆ AI ํ–‰๋™ ๋ณด๊ณ โ€ฆ ๋” ๋ฉด๋ฐ€ํ•œ ์ถ”์  ๋‹ค์ง

C1Sep 17
Huawei unveils new chip technology amid AI competition with Nvidia

Huawei unveils new chip technology amid AI competition with Nvidia

ํ™”์›จ์ด, ์—”๋น„๋””์•„ AI ๊ฒฝ์Ÿ ์† ์ƒˆ ์นฉ ๊ธฐ์ˆ  ๊ณต๊ฐœ

C1Sep 17
Tech Industry Split Over Calls to Coordinate AI Development

Tech Industry Split Over Calls to Coordinate AI Development

Tech Industry Split Over Calls to Coordinate AI Development

C1Sep 17
Global AI safety strategy hinges on US-China cooperation, yet each sees the other as the problem

Global AI safety strategy hinges on US-China cooperation, yet each sees the other as the problem

Global AI safety strategy hinges on US-China cooperation, yet each sees the other as the problem

C1Sep 17
Trump, downplaying the need to check AI development, says he won't cede advantage to China

Trump, downplaying the need to check AI development, says he won't cede advantage to China

Trump, downplaying the need to check AI development, says he won't cede advantage to China

C1Sep 14
Oprah Winfrey hopes you reach your 'AHA' moment at the Sphere

Oprah Winfrey hopes you reach your 'AHA' moment at the Sphere

์˜คํ”„๋ผ ์œˆํ”„๋ฆฌ๊ฐ€ ์Šคํ”ผ์–ด์—์„œ ๋‹น์‹ ์˜ 'AHA' ์ˆœ๊ฐ„ ๋„๋‹ฌ์„ ๋ฐ”๋ž€๋‹ค

C1Sep 14
New warning about AI risks reignites long-standing debate

New warning about AI risks reignites long-standing debate

AI ์œ„ํ—˜์„ฑ์— ๋Œ€ํ•œ ์ƒˆ๋กœ์šด ๊ฒฝ๊ณ , ์˜ค๋žœ ๋…ผ์Ÿ์„ ๋‹ค์‹œ ๋ถˆ๋Ÿฌ์ผ์œผํ‚ค๋‹ค

C1Sep 14