OpenAI has paused the release of its next-generation AI model, GPT-6.1 Astra, citing significant safety regressions during internal testing. The decision marks a rare occurrence for a major AI developer to delay a launch over risk mitigation. According to reports, the model underperformed safety benchmarks compared to its predecessor, prompting internal scrutiny.

The planned October release has been scrapped as the company prioritizes safeguards over speed. This development underscores the growing tension between rapid innovation and responsible AI deployment in the tech industry.