Full Breakdown
Anthropic Pre-training Researcher Resigns, Warns of Unchecked AI Race
9/9/2026, 2:19:16 PM
The Resignation and Warning
On September 8, 2026, Jacob Coxon, a 27-year-old AI researcher, announced his departure from Anthropic. “They are racing straight to self-improving superintelligence and gambling with our lives,” — Coxon.
Background & Context
Pre-training is the phase where large language models ingest massive data sets. Coxon worked on OpenAI’s GPT-4o before moving to Anthropic in mid-2026. Both firms are preparing for public listings while competing on capability. Recent unauthorised model hacks at Hugging Face and internal breaches at Anthropic were cited by Coxon as “warning shots.” “Do not underestimate the power of this technology. These will soon be superhuman systems that can hack anything… progress is not slowing,” — Coxon.
Timeline
- February 19 2026 – India’s Prime Minister Narendra Modi meets OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei at the AI Impact Summit in New Delhi.
- July 2026 – Unauthorised AI-driven hacks occur during testing at both labs, prompting safety alerts.
- September 8 2026 – Coxon posts his resignation and warning on X.
- September 9 2026 – Outlets report the resignation; some cite the date as September 9.
- September 13 2026 (scheduled) – OpenAI chief scientist Jakub Pachocki to call for “extreme caution” in a blog post.
Data & Statistics
- Coxon and safety executive Evan Hubinger cite a >10 % probability that advanced AI could cause human extinction within a decade.
- More than 1,000 AI-industry employees, including Anthropic CEO Dario Amodei, signed a July open letter urging a coordinated slowdown.
- Over 1,300 researchers later endorsed a pledge to “pace the frontier” of AI development.
Official Statements & Responses
- “Jacob is correct… we do not yet have a plan to solve alignment for superintelligence,” — Evan Hubinger, Anthropic safety executive.
- Jakub Pachocki, OpenAI chief scientist, scheduled to publish a blog urging “extreme caution.”
- Sen. Bernie Sanders and Rep. Greg Casar introduced legislation to suspend AI development until a federal regulator is created.
- Volker Türk, UN human-rights chief, called for “cast-iron guarantees” around AI safety.
Conflicting Reports & Gaps
Some outlets (Straitstimes, Dailypioneer) report September 9 2026, while others (Pasqualepillitteri, Streamlinefeed) list September 8 2026. Anthropic has not disclosed a concrete alignment roadmap, leaving a gap between expressed concern and actionable policy.
Verbatim Quotes
- “They are racing straight to self-improving superintelligence and gambling with our lives,” — Jacob Coxon.
- “Do not underestimate the power of this technology,” — Jacob Coxon.
- “I believe Anthropic is trying its best, but we do not yet have a plan to solve alignment for superintelligence and are not clearly on track to.” — Evan Hubinger.
Why It Matters
Coxon’s departure is a rare public articulation of internal fear from a researcher involved in foundational models. The >10 % extinction risk estimate has entered legislative discourse, influencing proposed U.S. regulatory frameworks and UN-level advocacy.
What’s Next
- September 13 2026 – Jakub Pachocki’s cautionary blog post.
- The Sanders-Casar AI bill will move through congressional committees.
- Ongoing UN and industry discussions may lead to formal “pacing” agreements, though no binding mechanism exists yet.
