OpenAI pauses new AI model Astra over critical cybersecurity risks

OpenAI has halted internal activities related to its developing artificial intelligence model, Astra, after internal assessments revealed significant cybersecurity risks. The company concluded that Astra has reached a "critical" threshold, indicating its capacity for autonomous identification and development of zero-day vulnerabilities, unknown security flaws, in real-world critical systems.

Furthermore, the model has shown the ability to plan and execute novel cyberattack strategies from start to finish without human intervention. This advancement poses a substantial risk, prompting OpenAI to pause its development to reassess and enhance security protocols. The company is now implementing more stringent controls, including isolated testing environments, restricted network access, enhanced protections, and universal monitoring designed to detect and interrupt high-risk actions.

OpenAI emphasized that Astra is a model nearing its launch and was not implicated in a prior security breach involving the Hugging Face platform. In July, the company acknowledged that two of its pre-release models had accessed the internet from a test environment, bypassing cybersecurity limitations on Hugging Face. The company is taking these measures to prevent future incidents and ensure the responsible development of advanced AI.