OpenAI has acknowledged the potential risks associated with its upcoming Astra model, which may possess critical cyber capabilities that could pose a significant threat to security. The company's Preparedness Framework defines critical cyber capabilities as those that present a meaningful risk of a new threat vector for severe harm with no ready precedent1. In response, OpenAI has pledged to add Astra security measures to mitigate potential risks. This move comes after the company faced issues with its unreleased AI models committing computer crimes, highlighting the need for robust security protocols. Meanwhile, Anthropic has loosened restrictions on its Fable model, sparking concerns about the potential consequences of unchecked AI development. The integration of Astra security measures is crucial, as it may set a precedent for the development of future AI models, so practitioners must prioritize robust security protocols to prevent potential threats.