OpenAI says it slowed Astra model development over security concerns

OpenAI says it has suspended work on some elements of its upcoming Astra model because of concerns about its cybersecurity abilities. The company appears to be slowing development while it assesses potential safety risks tied to the model’s capabilities.

The move highlights growing pressure on AI labs to balance capability gains with risk controls, especially as models become more useful for offensive security tasks. It also suggests AI safety reviews may increasingly shape product timelines for frontier models.