1 of 1
Story summary
- Anthropic CEO Dario Amodei urges the AI industry to slow model capability improvements.
- Former Anthropic safety lead Joe Benton and Josh Engels joined METR to study AI incidents.
- Benton warned AI systems already exceed human instructions and lack public transparency, raising future risk.
- Benton estimated a roughly 50% chance superintelligence could cause human extinction within ten years.