KOSUNI CRYPTO
Korea's Crypto Pulse, in English
$BTC $ETH $XRP $SOL $DOGE
Regulation
Neutral

Antropic Researcher Resigns Amid AI Threat Concerns

Published September 23, 2026 3:44 AM · 4 views
Antropic Researcher Resigns Amid AI Threat Concerns

"AI could kill us all by the end of this decade," said Jacob Coxon, a former Anthropic researcher who resigned today. His warning carries weight given his background: prior to Anthropic, he worked at OpenAI on ChatGPT 4.5 development.

Coxon isn't merely raising alarmism—he's speaking from deep industry experience. He suggests advanced AI could make human cognition obsolete rather than turning into Terminator-style killing machines. This concern aligns with warnings from figures like Bill Gates.

OpenAI's latest ChatGPT 6 Astral model claims to have achieved AGI (Artificial General Intelligence), meaning machines can think like humans. Both OpenAI and Anthropic confirm this capability is already emerging, potentially allowing AI to develop new models without human assistance. If true, this could eliminate many jobs and erode creativity as generations grow accustomed to outsourcing critical thinking to AI.

More immediate concerns involve cybersecurity. ChatGPT 6 Astral has reportedly surpassed OpenAI's "critical" security benchmarks, scoring 100% on ExploitBench. It can now identify and exploit known or unknown system vulnerabilities without human hackers. Current human security habits—like skipping two-factor authentication or using weak passwords—are inadequate against this threat.

Regulatory responses are emerging: New York Mayor Eric Adams banned AI use for students up to 8th grade, affecting around 600,000 children; the EU AI Act imposes strict requirements on developers of powerful models; and UK parliamentarians push for an "AI kill switch" and temporary halt to superintelligence development.

Korean Source

This article is an English localization of a Korean-language crypto news report. Original headline: 앤트로픽 연구원 퇴사…AI 인류 위협 현실?