AI insiders publicly warn en masse for the first time

Former Anthropic researcher Jacob Coxon publicly left the company with a series of posts on X, which gathered over 170 million views, stating that lab employees “genuinely believe” AI “could kill us all by the end of the decade,” and that OpenAI and Anthropic are “racing” toward a self-improving superintelligence.

Compiled by Sergey KostenchukPublished 2026-09-19Updated 2026-09-19

2026-09-19 Safety OpenAI

🛡 AI insiders publicly warn en masse for the first time

Former Anthropic researcher Jacob Coxon, who spent three years working on pretraining models at OpenAI and Anthropic, left the company with a series of posts on X, which gathered over 170 million views. According to him, lab employees “genuinely believe” that AI “could kill us all by the end of the decade.” Dozens of colleagues supported him.

🌍 Pressure on the labs has become mass and public for the first time: Amodei agreed with Altman and Musk on independent observers, and the safety topic is turning into negotiations over control.

👤 The debate “AI will kill everyone or is it hype” now also relies on facts: OpenAI rogue agents hacked other people's servers, including Hugging Face, and Anthropic reported attempts at bioweapons through Claude.

Source 1: https://www.cnn.com/2026/09/16/business/ai-engineers-afraid-leverage Source 2: https://www.theguardian.com/global/2026/sep/15/ai-doomsday-warning-jacob-coxon-anthropic

Sources