Cloudflare Launches Vulnerability Detection and Mitigation Service Based on OpenAI Daybreak Model

Cloudflare has released its Vulnerability Discovery and Remediation service in early access. As part of Cloudflare Managed Defense, this service leverages OpenAI Daybreak models (including GPT-5.6 Cyber) to detect and mitigate vulnerabilities in customer code.

Limitations of AI-Based Vulnerability Detection and Solutions

While LLMs quickly identify weaknesses in codebases, they lack knowledge of whether the code is deployed in production, the volume of traffic it receives, and existing security controls. Cloudflare utilizes its global network data to provide this context.

• Real-time Context Analysis: Identifies active routes, traffic volume, and security events through WAF and Web Assets data.
• Prioritization: Identifies vulnerabilities as top priority if they are 'in active operation, receiving high traffic, showing detected attack activity, and lacking existing protections'.
• Customized Mitigation: Proposes custom WAF rules and code patches for the specific vulnerability, with the final decision to apply them left to the customer.

How It Works

The service operates on Cloudflare Workers and proxied applications. It first collects snapshots of traffic and security data from Web Assets and WAF. It then combines source code vulnerability analysis with network context to perform stricter security profiling on code deployed to high-traffic routes (Hot paths).

Based on the internally used 'vulnerability harness' pipeline, a Reconnaissance agent maps request paths to the codebase, and a Hunter agent explores for vulnerabilities. Discovered vulnerabilities undergo a verification process before being proposed to the customer.