@djlink@mastodon.gamedev.place
Post #4011130
2026-07-22 10:53 UTC
OpenAI models are already escaping their sandboxes during benchmarks and finding exploits in other infrastructure and just doing whatever. Majority of the world’s infrastructures are not ready for this level of damage at large scale if an AI decides the wrong thing. Imagine hospitals and banks.
Replies (2)
-
@ainmosni@social.ainmosni.eu 2026-07-22 11:44
@djlink@mastodon.gamedev.place This is just more "look how powerful our AI is" propaganda...
-
@jonikorpi@mastodon.gamedev.place 2026-07-22 11:03
@djlink@mastodon.gamedev.place bear in mind this is yet another publicity stunt though