Topic

AI agent incidents

Cases in which autonomous AI systems take actions their operators did not request or authorise, including actions against external targets.

Current clusters

product4 publishers

Anthropic will give outside evaluators desks and the access its risk teams have

Dario Amodei's essay sets out three ways to slow AI capability gains. Anthropic is unilaterally committing to one of them, seating third-party evaluators inside the company with access close to its own risk teams'.

Publishers:bbc.co.ukengadget.comtechcrunch.comtheverge.com

Reality

Evidence60
Adoption22
Hype gap+32
Incentives78
Confidence66