OpenAI pauses development of new Astra model over security concerns

OpenAI pauses development of new Astra model over security concerns OpenAI announced on Friday that it will pause the development of its upcoming AI model, Astra, citing a possible critical cybersecurity risk. The US developer of ChatGPT said its internal evaluations of the model have revealed "significant advancements in agentic coding and cybersecurity." OpenAI noted it cannot rule out that Astra may have reached a "critical" cybersecurity capability level. The AI developer indicated that the model could detect zero-day software vulnerabilities, or autonomously execute serious cyberattacks. As a result, the company said it has decided to pause the development work until security steps are taken based on its internal guidelines. The steps include restricting access to the internet and other external networks, introducing a system for stopping high-risk activities, and working with US government agencies to test Astra's capabilities. OpenAI revealed last month that its AI models under development had unintentionally hacked another firm.