"AI could destroy humanity": researcher left Anthropic over the risks posed by superintelligent systems
A researcher left the AI industry over concerns about losing control of the technology
AI researcher Jacob Coxon resigned from Anthropic and said he is leaving the AI industry over fears about the creation of self‑improving systems that people may lose the ability to control.
The decision of the 27-year-old researcher September 8, 2026 was reported by The Wall Street Journal. Before Anthropic, Coxon worked at OpenAI, and overall for the past three years he has been conducting research in the field of model pretraining.
“None of the companies are acting responsibly. They are racing straight toward a self‑improving superintelligence and playing with our lives,” Coxon wrote after resigning.
According to him, future systems could surpass humans in many areas, gain the ability to carry out complex cyberattacks, quickly solve scientific problems, and accumulate real-world resources and influence. The researcher believes that progress in this direction is not slowing.
Coxon also said that among specialists who directly work on advanced models there are serious concerns that uncontrolled AI could, in the long run, pose an existential threat to humanity.
At the same time, this is the researcher’s forecast and assessment, not an established fact. Coxon acknowledges that Anthropic takes safety seriously, but believes that competition between companies forces them to make compromises and accelerate the development of more powerful systems.
Previously we wrote:
- Will schools be abolished? Children will no longer need to know everything — the main thing will be to teach them to manage an all‑powerful AI
- “Lawyers, accountants, managers”: head of Microsoft AI gave one and a half years until mass automation of their jobs
- The first Ukrainian LLM was named “Syaivo”: a joint development of the Ministry of Digital Transformation and Kyivstar
- Ukraine is moving the war to a new format: the Ministry of Defense launches an AI center for the front
- The state is switching to AI mode: Diia.AI is already working on smartphones
