OpenAI's Daybreak uses AI smarts to find security flaws

OpenAI has launched Daybreak, a cyber defense suite of tools that uses the intelligence of the company's LLMs to find and remedy software vulnerabilities.

The company hasn't freely released the tools to everyone; instead, it'll be "working with (...) industry and government partners" as it prepares to deploy "increasingly more cyber-capable models" in the future.

Mashable 101 Fan Fave: Vote for your favorite creator today.

According to OpenAI, Daybreak combines the intelligence of OpenAI models (in particular, the new GPT-5.5) with the extensibility of Codex (OpenAI's coding tool). Partners can use Daybreak for "secure code review, threat modeling, patch validation, dependency risk analysis, detection, and remediation guidance."

Three models are offered at this time. The default GPT-5.5 model can be used for general purpose work. GPT-5.5 with Trusted Access For Cyber is for "most defensive security workflows," which includes secure code review, malware analysis, and patch validation. Finally, GPT-5.5-Cyber can be used for authorized red teaming, penetration testing, and controlled validation.

If this sounds familiar, you're thinking of Anthropic's Project Glasswing initiative and Mythos, an AI that's so good at finding software vulnerabilities that Anthropic has only given access to selected partners.

OpenAI did not publicly disclose pricing for Daybreak; those interested can contact the company's sales team for a quote. Several companies are listed as partners, including Cloudflare, Cisco, Oracle, and Akamai.