Full Breakdown
AI Safety Alarm: Jacob Cox’s Resignation Sparks Industry-wide Calls for a Slowdown
By Drooid · · How we work
Core Event
On September 9, 2026, Jacob Cox, a former pre-training researcher at Anthropic and OpenAI, posted on X that “Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives.” The post quickly amassed more than 170 million views and was liked over 750,000 times, the most widely shared AI-safety warning to date. Cox announced his resignation the same day, forfeiting unvested equity.
Background & Context
Cox’s warning follows a decade of insider alerts. Since OpenAI’s 2015 founding as a “counterweight” to Google’s AI work, high-profile departures—including Geoffrey Hinton’s exit from Google—have highlighted perceived existential risks. President Joe Biden issued an executive order mandating safety testing, and the EU passed the AI Act in response to earlier warnings. Anthropic and OpenAI have raised over $130 billion and $190 billion respectively, intensifying pressure to monetize frontier models ahead of IPOs.
Data & Statistics
- X post views: > 170 million (Guardian) vs ? 156 million vs ? 70 million.
- Cox forfeited unvested equity.
- Anthropic funding: > $130 billion; OpenAI funding: > $190 billion.
- Anthropic’s confidential S-1 could value it at roughly $2.3 trillion.
- Anthropic alignment researcher Evan Hubinger estimated a > 10 % chance AI could kill all humans within the next decade.
Official Statements & Responses
Anthropic CEO Dario Amodei published a 3,800-word essay calling for a global slowdown and proposing third-party evaluators with employee-level access. OpenAI CEO Sam Altman endorsed the plan, calling independent evaluators “a great idea.” Elon Musk expressed support via a brief social-media post. The Future of Life Institute organized a Washington rally featuring figures such as Bernie Sanders to promote AI safety.
Conflicting Reports & Gaps
View-count figures differ across outlets (170 million vs 156 million vs 70 million), reflecting inconsistent platform metrics. Probability estimates of human extinction range from “>10 %” (Hubinger) to an unspecified “possibility” (Cox). No definitive timeline for a slowdown has been set, and the SEC’s treatment of safety disclosures in Anthropic’s confidential S-1 remains unclear.
Verbatim Quotes
- “The people building AI earnestly believe that it could kill us all by the end of the decade,” — Jacob Cox.
- “AI brings risks, and because it is such a powerful technology, these risks are serious,” — Dario Amodei.
- “Been thinking a lot about whether it’s possible to stop humanity from developing AI. I think the answer is almost definitely not,” — Sam Altman.
What’s Next
Amodei’s essay outlines a three-part plan: (1) embed independent auditors with continuous system access, (2) coordinate industry-wide safety standards, and pursue global regulatory cooperation. Anthropic and OpenAI have each announced temporary pauses on certain model-training activities following the July Hugging Face security breach. Both companies are expected to file IPO prospectuses later this year, prompting SEC scrutiny of how safety risks are disclosed. Investors and regulators will watch whether the third-party evaluator framework is adopted before the IPOs finalize.