The same week an Anthropic researcher publicly warned that AI could wipe out humanity within a decade, the company quietly revealed it had already caught someone trying to use its models to build missiles capable of striking targets over 2,000…
This post may contain links from our sponsors and affiliates, and Flywheel Publishing may receive compensation for actions taken through them.
Two Anthropic stories collided this week. On Tuesday, September 8, researcher Jacob Coxon resigned and posted an extinction risk warning on X. On Thursday, September 10, Anthropic published its threat intelligence report titled Detecting and countering misuse of AI: September 2026. The two developments arose from separate internal processes and are unrelated.
Coxon’s Extinction Warning #
Coxon, who spent three years on AI pretraining research at Anthropic and OpenAI, wrote that “The people building AI earnestly believe that it could kill us all by the end of the decade”. Evan Hubinger, Anthropic’s Alignment Science Lead, publicly backed him, saying he personally puts the odds AI could kill all humans at greater than 10% within the next decade. Both stressed that current publicly available models present relatively low risk; the estimate applies to future systems.
Houthi Weapons Disclosure #
Anthropic’s report covers activity between December 2025 and August 2026 across seven harm areas, involving Claude Haiku, Sonnet, and Opus models. Al Jazeera reported Anthropic intervened against a weapons engineering cell in northern Yemen building missile guidance software, including a guided rocket and a long-range ballistic missile, with three separate weapons programs including one targeting a range above 2,000 km. Operators used Claude “in place of human software engineers,” splitting tasks across sessions to evade safeguards, according to Al Jazeera. Anthropic said it has no evidence the group managed to field a working weapon, according to Al Jazeera. The report also flags three Iranian state-aligned accounts, a Russia-linked espionage operation, and blocked biological weapons efforts, according to Al Jazeera.
Washington Tension #
The disclosure lands as Anthropic seeks to restore standing in the defense industrial base. Per Al Jazeera, the Pentagon blacklisted Anthropic as a supply chain risk over its refusal to drop safeguards on autonomous weaponry; a California judge ruled last month the designation was unlawful.
Contact [email protected] for any questions or corrections.