Jacob Coxon
Former pretraining researcher at OpenAI and Anthropic
British AI researcher who worked on model pretraining at OpenAI and then Anthropic. He resigned from Anthropic on Sept. 8, 2026, posting that neither company was acting responsibly and that they were "gambling with our lives" by racing toward self-improving superintelligence.
Key arguments & positions
- Argues frontier labs are racing toward self-improving superintelligence without adequate control, and that neither OpenAI nor Anthropic is acting responsibly.
- Has said today's AI tools are safe for everyday use; his concern is where the pace of capability gains leads.
Accomplishments
- Pretraining researcher at OpenAI (from 2023, including work on GPT-4o) and Anthropic (from 2024).
- Represented the U.K. at the International Mathematical Olympiad in 2016 and 2017; studied mathematics at Cambridge.
Papers & key writings
His views appear in his Sept. 8, 2026 resignation post on X and in interviews, not long-form writing.
Links
In the library
Nothing of theirs is filed in the hub yet. Browse the full library.
Recommended next
Hand-picked from the hub based on what Jacob Coxon covers.
These Are the Most Urgent AI Risks, According to 272 Experts
MIT Sloan summary of research surveying 272 experts on which AI risks could cause the most harm in the next five years.
Why this: Covers AI safety and AI risk too
AISafety.info
Free question-and-answer site explaining AI risk arguments in plain language, founded by Rob Miles and maintained by volunteers. Answers are organised as linked questions from beginner to advanced, covering how AI is advancing, why systems may pursue goals, alignment research and AI governance. Includes Stampy, a chatbot that answers AI safety questions with sources. Open source on GitHub; run as a project of Ashgro Inc, a US 501(c)(3) charity.
Why this: Covers AI safety and AI risk too
PyRIT
Microsoft's open-source toolkit for red-teaming AI systems: automated attack prompts, scoring of the responses, and repeatable runs. Free.
Why this: Covers AI safety and AI risk too
Concrete Problems in AI Safety
The 2016 paper that framed AI safety as a set of specific engineering problems — side effects, reward hacking, unsafe exploration — rather than a philosophical worry. Free on arXiv.
Why this: Covers AI safety and AI risk too