Cloudflare opens early access to context-aware vulnerability reviews

Cloudflare announced early access to Vulnerability Discovery and Remediation on 3 September through Managed Defense. The service analyses code that a customer authorises it to inspect, using OpenAI Daybreak models including GPT-5.6 Cyber. It combines findings with available route, traffic and security-control information to help establish which weaknesses are relevant to a deployed application.

Cloudflare says the workflow validates findings and proposed patches or web-application firewall rules before presenting them for review. Models cannot apply those changes themselves. The service is invitation-only and begins with a bounded application investigation, rather than unrestricted access to a customer’s environment. Its purpose is to connect source-level findings to evidence of exposure, so a team can assess a proposed fix in the context of the system it actually runs.