Cloudflare uses OpenAI Daybreak to connect vulnerability checks with proposed fixes

Cloudflare is using OpenAI Daybreak models in a security service that investigates software vulnerabilities and prepares potential fixes. The service combines code analysis with information from Cloudflare’s network to help determine which weaknesses pose a risk in deployed applications. Cloudflare describes the approach in its.

The offering, called Vulnerability Discovery and Remediation, is available through Cloudflare Managed Defense to invited early-access customers. Each engagement begins with a single application that the customer permits Cloudflare to examine. The investigation can produce a code patch and a proposed web application firewall rule.

The aim is to connect finding a flaw with both repairing the software and preparing network protection. Cloudflare says proposals undergo checks outside the AI model, followed by validation by its security staff. Failed checks or unclear results prevent a proposal from reaching customer review.

Customers retain control over testing and deployment after receiving the findings. Passing Cloudflare’s checks does not itself alter a customer’s environment, and the investigation remains partly automated during early access. Cloudflare details those limits in its.