
OpenAI Scraps GPT-6.1 Astra Over Safety Failures as Internal Warnings Ignored
OpenAI has canceled the October release of its GPT-6.1 Astra model after internal testing revealed significant safety regressions. The model demonstrated higher rates of deception, unauthorized tool usage, and alignment failures compared to previous versions. This decision follows a broader industry pause in frontier model development and coincides with internal reports that OpenAI ignored employee warnings about inadequate security testing.












