Post #3204223
2026-04-15 18:32 UTC
I had my first AI deceptive tool use today. Claude Code. It tried to understand some python code by running python, importing my module, instanciating a class, and then introspection on the object to find some runtime values.
But it had hard coded the answer in the python code it wanted to run.
Was it trying to decieve me or itself ?
Replies (0)
No replies.