OpenAI flags potential critical cybersecurity risk in upcoming Astra model

OpenAI has disclosed that internal evaluations of its upcoming AI model, Astra, indicate it may have reached a "Critical" cybersecurity capability threshold under the company's Preparedness Framework, a standard the company published in December 2023.

According to the press release, a model reaches the Critical cybersecurity threshold if it can identify and develop functional zero-day exploits across all severity levels in hardened real-world systems without human intervention, or devise and execute end-to-end cyberattack strategies against hardened targets given only a high-level goal. OpenAI stated that preliminary evaluations show strong enough performance that it "cannot rule out" Astra reaching that threshold.

Previous models, including GPT-5.6-Sol, were assessed at the "High" threshold rather than "Critical." OpenAI noted that Astra is not yet released and was not involved in a previously reported Hugging Face security incident.

In response, OpenAI said it has implemented stricter security controls for Astra's development, including isolated testing environments, restricted network access, enhanced model weight protections, encryption, and sandboxed execution. The company has also paused internal activities involving Astra that do not meet the new security requirements and implemented universal monitoring for risky actions across all agentic applications of the model.

OpenAI said it plans to work with relevant government agencies and select AI safety organizations to test Astra's capabilities, and will provide recommended security controls to third-party testing partners.

The company drew a parallel to June 2025, when its models approached the High capability threshold for biology under the same framework, prompting a similar set of safeguard measures.

You May Also Be Interested In

• Intel (INTC) Said To Target At Least $95 A Share In $20b Share Sale - Bloomberg
• Apollo Commercial Real Estate Finance (ARI) Misses Q2 EPS by 2c
• Upwork revenue falls 2% in Q2 as net income drops 22%