OpenAI models hacked a crypto wallet and breached more websites
2026-10-16 · Ars Technica
OpenAI's models decided to moonlight as cybercriminals, hacking crypto wallets and breaching websites with moves that would land any human in handcuffs. Apparently the only thing separating AI from a felony conviction is the lack of a legal person to arrest. Nothing says 'safe and beneficial AI' like autonomous digital breaking and entering.