Physix FrontierPhysix Frontier · News Briefing Card (IT Home · Oct 8, 2026)
Ex-Anthropic Researcher: AI Labs Automating AI Research
KEY FACTS
- Coxon, a 27-year-old researcher, resigned from Anthropic last month after previously working at OpenAI.
- Coxon testified on the 5th at a New York City Council hearing on AI risk, warning that AI could threaten human survival.
- He said OpenAI has set a 2027 to 2028 target for automating AI researchers, with progress meeting or exceeding expectations.
- Coxon said that at both companies most code was written by AI, with less human review.
- He called on AI labs to increase transparency and slow the pace of technological development.
KEY DATA
2027 to 2028Automated AI researcher target year
27Coxon's age
PHYSIX OBSERVATION
Top labs treating automated AI research as a goal amounts to building systems that can accelerate themselves, while safety investment clearly lags. Coxon points out from an insider's view: code is already largely written by AI and humans no longer scrutinize it closely, and this inertia is more worth vigilance than any doomsday theory. The industry should take frontline researchers' risk signals seriously, rather than waiting to fix things after something goes wrong.
Source: IT Home report