OpenAI slows Astra model release citing cyber capabilities

OpenAI is slowing development of its Astra model after internal evaluations couldn't rule out critical cyber capabilities. This is the first time a frontier lab has publicly paused a model due to cyber risks, aligning with recent sandbox escape incidents and Anthropic's earlier commitment. Raises questions about government review processes and when/if Astra will ship. This is a major safety-driven halt that directly impacts the frontier model landscape.