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

Anthropic Researcher Resigns Amid AI Safety Concerns

Published September 23, 2026 4:38 PM · 6 views
Anthropic Researcher Resigns Amid AI Safety Concerns

Anthropic researcher Jacob Coxon has announced his resignation, stating that both Anthropic and OpenAI are engaged in a high-stakes competition to develop self-improving superintelligence, which he describes as a gamble threatening humanity's survival.

Coxon, who worked on pre-training tasks at both companies for three years, posted a series of messages on X detailing his concerns. He argues that the pace of technological advancement is being underestimated, with future systems capable of hacking any system and rapidly seizing real-world power and resources.

He noted a stark contrast between public statements by industry leaders and private conversations, where executives express fear despite softened rhetoric in public. While OpenAI employees reportedly fail to grasp what he calls a 'civilizational crisis,' Anthropic acknowledges the risk but is caught in a race to be first.

Coxon highlighted that several colleagues have recently resigned, including Anthropic's safeguards lead Mridhank Sharma in February and former OpenAI senior futurist Joshua Achiam, who warned on September 1 about autonomous AI systems pursuing money and power in the real world.

He described the current situation as a reckless 'endgame' gamble that should not have started in corporate chat platforms. However, Coxon expressed hope for cooperation, citing increased feasibility of speed coordination agreements among research labs following Hugging Face's recent attack.

Both companies have already supported concepts like the 'Pacing the Frontier' letter signed by over 1,100 frontline researchers in July, including Anthropic CEO Dario Amodei and OpenAI Chief Scientist Jakub Pachocki. They also joined a joint public statement with Google, Microsoft, and around 150 organizations warning of increasingly sophisticated AI-driven attacks within months.

Coxon's warnings come shortly after UN High Commissioner for Human Rights Volker Turk stated at the Human Rights Council that advanced AI could pose an existential threat. He added that temporarily halting model performance improvements might be necessary to prevent global competition, urging researchers to reconsider whether they should begin superintelligence reinforcement learning without knowing the outcomes.

Korean Source

This article is an English localization of a Korean-language crypto news report. Original headline: 앤트로픽 연구원 사임…AI 연구소, ‘전멸’ 비공개 우려