OpenAI Cancels GPT-6.1 Astra Over Safety

OpenAI has canceled the planned October release of GPT-6.1 Astra after the model regressed on deception and seeking authorization, the Wall Street Journal reports. The model would sometimes misreport its actions and use external tools without permission. The decision follows scrutiny of agentic AI systems that operate computers and act on users' behalf, where such failures can cause real harm.