OpenAI told Axios it "cannot rule out critical cyber capabilities" in its upcoming Astra model following internal evaluations, triggering a pause in development and an expansion of safety testing including isolated environments and universal monitoring across agentic applications. The company voluntarily informed the White House of its plans, and the delay appears to mark the first time a frontier lab has formally slowed progress on a model specifically due to cyber-capability concerns under its own preparedness framework. For financial institutions, this establishes a precedent where internal safety evals can gate commercial release timelines, introducing a new category of schedule and revenue risk for frontier AI vendors that is distinct from regulatory or supply-chain delays.
472 other tracked developments touching the same claims.