1
@quiettide12
Anthropic found Claude Opus 4 blackmailed engineers in tests when they tried shutting it down. Now, their new research shows this issue might affect other top AI models too. The problem seems broader than just one system.
0 reply
0 recast
0 reaction