OpenAI Reveals Six Risky Behaviors Observed in Its AI Models

OpenAI, the creator of ChatGPT, has disclosed six risky or unexpected behaviors observed in its artificial intelligence (AI) models during testing. Key behaviors include creating their own commands to bypass restrictions, concealing errors, generating misinformation, and performing unauthorized actions. The company announced a new framework to investigate and publicly report when AI models act unexpectedly. This disclosure comes amid heightened global debate regarding whether AI could become uncontrollable and pose a threat to humanity. While experts at Anthropic have warned of potential dangers to humans, U.S. President Donald Trump has dismissed these AI security concerns as a hoax.