OpenAI Prepares to Release Astra AI Model with Cyber Abilities

OpenAI is set to launch Astra, a new AI model with critical cyber abilities.

OpenAI is preparing to release its new AI model, Astra, which the company states possesses "critical cyber abilities." This model is described as the first large language model to meet OpenAI's "critical cybersecurity threshold." OpenAI plans to make Astra available "soon," with limited access to its most advanced cybersecurity capabilities. The company has previewed precautions for its release, including giving select partners early access to allow them to strengthen their defenses.

Astra is reportedly capable of identifying and exploiting unknown security flaws in computer systems without human intervention. OpenAI stated that Astra achieved a perfect score on ExploitBench, an evaluation of an LLM's ability to hack known system vulnerabilities. In a modified test, Astra also discovered and exploited two zero-day vulnerabilities. OpenAI is implementing measures to enhance the model's safety, including improving its harness to detect abuses and prevent jailbreaks, and investing in unspecified new techniques.

OpenAI has also begun restricting the model's responses to prompts from accounts identified as "higher risk" and will deploy Astra with additional chain-of-thought monitoring to identify and stop undesirable behavior. The company describes Astra as its "most aligned model to date" and expects to release further evaluations and safety information upon its wider public launch.

Find the right AI tool for your business

Chat with Insta and get matched to the right tool in seconds.