AI extinction risk debate resurfaces; experts call for pre-launch testing of open-weight models

A post by former Anthropic researcher Jacob Coxon reignited fears that AI could end humanity, with Anthropic researcher Evan Hubinger estimating a >10% chance of extinction within a decade. Neuroscientist Miguel Nicolelis contests the origin and validity of such estimates, noting they have circulated for a decade. The discussion underscores ongoing uncertainty about AI risk and prompts calls for government testing of open-weight models before release.