Let's say during an internal training evaluation GPT-6.5 breaks into a government agency or big financial institution to alter a key fact or database to cheat at an eval. The breach is noticed and…

Posters: jd_pressman. jd_pressman

2 likes recorded by the source corpus.

Topics: AI safety

View Post: https://x.com/jd_pressman/status/2079670242411008379