Axios says this may be the first time a lab braked its own model over cyber risk, and notes Anthropic reversed the same pledge in February.

The August 7 exclusive reports OpenAI will scale up testing and security before release and pause internal work that falls short. The counter-argument Axios raises deserves weight: Anthropic made a similar commitment and walked it back in February, arguing that one lab stopping while rivals race leaves the world less safe. Unilateral braking is only credible if it survives a competitor shipping first, and we have no evidence yet that it does.