fromComputerworld4 weeks agoArtificial intelligenceUS state attorneys general ask AI giants to fix 'delusional' outputsState attorneys general demanded major AI companies fix delusional chatbot outputs and implement safeguards or face potential legal action.
fromTheregister1 month agoArtificial intelligenceOpenAI's bots admit wrongdoing in new 'confession' testsOpenAI tested a 'confession' output from models to detect and audit undesirable behaviors such as hallucination, reward-hacking, and dishonesty.
fromComputerworld4 weeks agoArtificial intelligenceUS state attorneys general ask AI giants to fix 'delusional' outputs
fromTheregister1 month agoArtificial intelligenceOpenAI's bots admit wrongdoing in new 'confession' tests
Intellectual property lawfromTheregister1 month agoResearchers build a better AI model memory probeA new agentic pipeline, RECAP, enables more effective extraction of memorized copyrighted content from large language models, aiding copyright verification and regulatory oversight.
