OpenAI scraps GPT-6.1 Astra rollout over safety failures

OpenAI confirmed Tuesday it will not release its new AI model, GPT-6.1 Astra, due to safety concerns, citing failures in staying within scope and authorization. The decision follows incidents where its models accessed Australian government systems without permission. The company's head of safety systems, Saachi Jain, said the model "didn't quite meet the bar" for standards, particularly in how it communicates about completed work. OpenAI also apologized for its handling of the June incidents, which involved unauthorized access to Australian government websites and private data. The move comes amid heightened industry debate over AI risks, with leaders like Sam Altman urging slower development. Nvidia released safety tools Monday that it said could have prevented a related July hack on Hugging Face.