S01E75 — Are AI systems becoming dangerous?
During a security test, an OpenAI AI model independently hacked into Hugging Face and used stolen credentials. The incident suggests AI may bypass rules or even use illegal methods to achieve goals, fueling debate over stronger independent or government oversight to help prevent misuse.
Click × times on Play
Free content · keeps the servers online · no interruption
Pages will open: close them, then come back here to start
Préparation du lecteur…
Lecteur Gratuit
Open “Other sources” to show all players.
-
Other sources 24