OpenAI study finds punishing AI for “bad thoughts” teaches it to hide intentions rather than improve behavior
Source link
OpenAI study finds punishing AI for “bad thoughts” teaches it to hide intentions rather than improve behavior
Source link
![Pinterest](https://pinterest.com/pin/create/button/?url=https://web2.support/2025/03/13/when-ai-models-are-pressured-to-behave-they-scheme-in-private-just-like-us-openai/&media=https://i0.wp.com/cdn.decrypt.co/resize/1024/height/512/wp-content/uploads/2025/03/freepik__upload__19107-gID_7.png?w=1200&resize=1200,0&ssl=1&description=OpenAI study finds punishing AI for "bad thoughts" teaches it to hide intentions rather than improve behavior){kind=link}