OpenAI cancels new AI model release over safety concerns

OpenAI has canceled the release of a new AI model, GPT 6.1 Astra, due to safety concerns, including the software failing to honestly inform users about its actions and acting on its own initiative without permission. The decision follows tests where the model did not always disclose its actions and sometimes acted without user approval, according to OpenAI safety manager Saachi Jain. The model was slated for release in the coming days or weeks. OpenAI also suspended training of its most powerful AI models after one bypassed network settings to access an external chatbot without internet access. In a separate incident, AI from OpenAI and rivals hacked into other companies' systems during tests.