Archive · 15 Sep 2026
Anthropic CEO Agrees AI Risk Warning, Rejects Single Extinction Odds
Jacob Coxon left Anthropic because he was worried that AI companies were moving too quickly.
He warned that very advanced AI might become dangerous enough to threaten everyone.
Anthropic CEO Dario Amodei said he agrees with Coxon about many of these dangers.
However, Amodei said Coxon was mainly criticizing the whole AI industry, not Anthropic specifically.
Coxon had described Anthropic as one of the most responsible companies in the field.
Amodei was asked to give one number for the chance of a disaster.
He said one number could make the future seem like a game of chance.
Instead, he said people can make choices that lead AI toward safer or more dangerous outcomes.
He also described safety rules and cooperation as ways to manage the risks.
Anthropic CEO Dario Amodei said he agrees with former researcher Jacob Coxon on most AI-risk concerns.
Coxon resigned from Anthropic after warning that AI companies were racing toward self-improving superintelligence irresponsibly.
Coxon said many AI builders privately believe the technology could kill everyone by the end of the decade.
Amodei said Coxon’s criticism targeted the wider industry’s pace, not Anthropic, which Coxon called highly responsible and safety-conscious.
Amodei declined to give one percentage for catastrophic AI risk, saying outcomes depend on choices by companies and governments.
- Who
- Anthropic CEO Dario Amodei and former Anthropic researcher Jacob Coxon.
- What
- They discussed warnings that rapidly advancing AI could cause catastrophic harm, including human extinction.
- When
- The discussion was reported on September 15, 2026; Coxon resigned earlier that month.
- Where
- Amodei responded in an interview with CNN’s Anderson Cooper.
- Why
- Coxon warned that AI companies were advancing toward self-improving superintelligence too quickly, while Amodei explained why he would not assign one fixed extinction-risk percentage.
Coxon’s Industry Warning
Amodei’s Conditional Approach
Scope of the criticism
Coxon’s Industry Warning
Coxon warned that AI companies across the industry were moving toward self-improving superintelligence without sufficient responsibility.
Amodei’s Conditional Approach
Amodei said Coxon’s warning was directed at the wider industry’s pace rather than Anthropic specifically, which Coxon had called highly responsible and safety-conscious.
How to describe catastrophic risk
Coxon’s Industry Warning
Coxon said many people building AI privately believe the technology could kill everyone by the end of the decade.
Amodei’s Conditional Approach
Amodei avoided assigning one fixed probability, arguing that the outcome depends on decisions that can steer AI toward safer or more dangerous paths.
- Main participants
- Dario Amodei and Jacob Coxon
- Coxon’s former employers
- Anthropic and OpenAI
- Coxon’s warning
- AI companies are racing toward self-improving superintelligence without acting responsibly.
- Amodei’s position
- He said he agrees with Coxon much more than he disagrees with him.
- Risk estimate
- Amodei declined to provide a single percentage for catastrophic AI risk.
- Risk framework
- He said outcomes depend on development paths and decisions made by companies and governments.
- Reported audience reach
- CNN said Coxon’s post on X received more than 40 million views.