OpenAI said its Astra model is the first from the company to meet the “Critical” cybersecurity capability threshold under its Preparedness Framework, according to a blog post published Sept. 1, 2026. The company said Astra will be released with stronger safeguards designed to address the risks associated with advanced cybersecurity capabilities. The announcement did not provide further technical details in the email. OpenAI’s Preparedness Framework evaluates the potential risks posed by frontier models and guides the safety measures required before release. The full announcement was published on OpenAI’s website. Post navigation Samsung Showcases Love-Inspired Design