OpenAI says it will limit access to the most advanced cybersecurity capabilities in its new AI model, Astra, as it prepares for rollout. The company indicates the restricted features are among the model’s most powerful tools for security-related tasks.
According to reporting across outlets, OpenAI’s decision is tied to internal assessment of Astra’s risk profile. The Wall Street Journal reports that internal testing rated the model as posing a “critical” cyber risk and found it could perform complex cyberattacks with limited human input, leading OpenAI to add additional safeguards. Bloomberg similarly frames the move as an effort to control access to Astra’s cutting-edge cybersecurity functionality.
While the outlets focus on the same core announcement, they differ in emphasis: Bloomberg highlights the access-limiting aspect of the rollout, while the Wall Street Journal underscores the internal findings and the “critical” rating that motivated extra security measures. Seeking Alpha’s coverage aligns with the overall theme that OpenAI is restricting the most advanced cybersecurity features, though it provides fewer specific details in the provided excerpt.
