# Ex-Anthropic researcher warns of an AI race

> Jacob Coxon says Anthropic is helping speed a race toward self-improving AI that its leaders believe cannot be avoided.
- Title: Ex-Anthropic researcher warns of an AI race · Business Insider
- Summary: Jacob Coxon says Anthropic is helping speed a race toward self-improving AI that its leaders believe cannot be avoided. Coxon's core claim is that treating…
- Keywords: ai safety, anthropic, openai, superintelligence, tech policy, business, technology, Ex-Anthropic, researcher, warns, race, Business Insider
- Source: Business Insider — https://businessinsider.com/jacob-coxon-explains-disagreements-anthropic-leadership-warning-2026-9
- Author: Brent D. Griffiths
- Published: 2026-09-15T21:34:48.223+00:00
- Read time: 1 min
- Topics: ai safety, anthropic, openai, superintelligence, tech policy, business, technology
## Coxon breaks ranks
Former Anthropic researcher Jacob Coxon used an X AMA to spell out why he thinks the company is wrong to keep pushing the frontier of AI.

> If the race is inevitable you should not contribute
## The fight is inevitability
Coxon says Anthropic leaders think a race to advanced AI cannot be negotiated away, so they feel forced to compete hard.

> They don’t believe it will be possible to negotiate.
## RSI is the flashpoint
His sharpest concern is recursive self-improvement, where AI systems could improve themselves and build stronger successors.

> They largely initiated the recent race to RSI
## The warning went viral
Coxon drew attention after resigning and accusing both Anthropic and OpenAI of racing toward self-improving superintelligence.

> gambling with our lives
## Amodei fears bad actors
Anthropic CEO Dario Amodei has argued that if his company slows too much, less safety-minded companies or governments could control the technology.

> the wrong people will be in charge of the technology
## Leaders now urge pacing
Amodei has called on frontier labs to slow capability gains and spend the extra time on alignment. Altman, Musk, and Hassabis backed the broad point.

> We must slow the pace
## Key takeaway

Coxon's core claim is that treating an AI race as inevitable can become the very thing that makes it happen.