OpenAI shelves GPT-6.1 Astra over safety concerns, halts October release

Model reportedly struggled with staying within authorised scope and accurately communicating its actions, even as rival Anthropic warns investors of potentially catastrophic AI risks