CLOUDFLARE has announced an early access offering called Vulnerability Discovery and Remediation (VDR), now part of Cloudflare Managed Defense. The service aims to help organisations prioritise and remediate vulnerabilities by combining real‑world context from production traffic and security events with automated vulnerability analysis. Using OpenAI Daybreak models, including GPT-5.6 Cyber, the system conducts reconnaissance, hunting, and validation against codebases authorised by the customer.
If a vulnerability is detected, Cloudflare proposes patches and mitigations, but the customer remains in control and decides whether to implement them after review.
The approach integrates traffic and security data from Web Assets and the Web Application Firewall (WAF) to add production context to findings. For Cloudflare Workers and proxied applications, the vulnerability harness maps findings to live routes, pulls route‑level data, and, where appropriate, suggests Cloudflare WAF custom rules to reduce exposure while code changes are reviewed. Proposals pass through multiple checks before presentation, and all patches or rules are validated by Cloudflare engineers.
The service is invitation‑based during this early access phase, with engagements starting on a single authorised codebase per customer. The article notes that this model‑backed, context‑driven workflow is designed to prioritise meaningful, production‑relevant fixes rather than generic alerts.