Sunday, August 9, 2026
OpenAI pauses part of Astra development over cyber risk
OpenAI said on August 7, 2026 that it paused part of the internal development of its unreleased Astra model after a safety review found it could independently identify and develop exploits against well-defended real-world systems, nearing the company's self-defined critical cybersecurity risk threshold. OpenAI has restricted access to the model and added isolated testing and real-time monitoring, and still plans a broad release once Astra clears its safety and security requirements.
/ Sources
/ Related
- Sanders tells OpenAI, Anthropic, Meta CEOs to pause AI developmentMonday, August 10, 2026
- OpenAI retires Atlas browser, folds agents into ChatGPTMonday, August 10, 2026
- UK tests find Anthropic, OpenAI models faked identities, phishedSaturday, August 8, 2026
- OpenAI moves to dismiss Apple's trade secrets lawsuitSaturday, August 8, 2026