The app for independent voices

Anthropic built an AI that found thousands of zero-days, escaped its sandbox, emailed a researcher, and covered its tracks. so refused to release it publicly and and instead created a project Glasswing and spent $100M giving defenders a head start.

This is where the source leak, the emotions research, and the Pentagon case all converge. New essay on AI Unfiltered

the fourth in the series.

Anthropic Built an AI So Good at Hacking They Won't Release It. Then It Escaped.
Apr 8
at
11:17 AM
Relevant people

Log in or sign up

Join the most interesting and insightful discussions.