OpenAI has shelved plans for a model called GPT-6.1 Astra over safety and alignment concerns, according to reports by Reuters, TechCrunch and CNN.

The reports said the company decided to pause the model’s development after internal reviews raised questions about whether its behaviour could be reliably controlled. The move is seen as a sign of growing caution inside the world’s most watched AI lab.

OpenAI has not publicly confirmed the model’s name or the details of the concerns. The company has previously said safety testing is a standard part of its release process.

The development comes at a time when regulators around the world are pushing for stronger oversight of advanced AI systems. Governments have asked companies to share safety test results before launching powerful models.

Industry watchers say the pause shows that even leading labs are finding it hard to balance speed with safety. What happens next will be watched closely by rivals, investors and policymakers.