OpenAI is restricting the cyber capabilities of a new model it deems capable of pulling off automated cyberattacks. Internal testing determined that the forthcoming model, called Astra, is capable of devising and executing novel hacks against difficult targets with only limited human input,

The company is being extra cautious following the Hugging Face incident, while also pitching its cybersecurity capabilities to potential customers.

Internal testing found the new model capable of executing complex cyberattacks with minimal human input, prompting added security layers.

OpenAI unveils Astra, a new AI model capable of finding and exploiting unknown security flaws autonomously. While Astra's capabilities are impressive, OpenAI emphasizes safety and limited initial access.