Cloudflare has integrated OpenAI models accessed through its Daybreak platform into a new early-access service called Vulnerability Discovery and Remediation, marking a significant step in linking SI-driven code analysis directly to network defense. The initiative aims to automate the identification of security flaws in customer-authorized code while leveraging Cloudflare’s global network data to prioritize threats.
What Happened
Cloudflare, which proxies approximately 20% of web traffic, is combining its live traffic data and security signals with OpenAI models to investigate code and assess potential vulnerabilities. The new service, offered through Cloudflare Managed Defense, uses these SI tools to identify issues in source code, validate findings against production activity, and prepare proposed mitigations. These mitigations include code patches and custom web application firewall rules, which are reviewed by Cloudflare’s team before being shared with customers for final implementation.
The integration also extends to Cloudflare’s internal engineering practices through the use of Codex, an SI tool from OpenAI. Mark Anderson, President of Revenue at Cloudflare, stated that 100% of the company’s software developers use Codex to boost output and improve code quality. Anderson highlighted a recent internal project where a team of four developers, aided by Codex, completed a rebuild in three days that would have previously required months and ten times the manpower, costing only $1,100 in tokens.
Why It Matters
For security teams overwhelmed by the volume of vulnerability alerts, this integration offers a method to contextualize code flaws with real-time production data. By connecting static code analysis with dynamic network signals, the service helps determine if an affected route is active or already protected, allowing human engineers to focus their judgment on critical decisions rather than manual data sifting. Anderson noted that the automation allows machines to handle grunt work, freeing humans to exercise judgment on important engineering and security choices.
The Bottom Line
Cloudflare’s deployment of OpenAI models via Daybreak and Codex illustrates how major infrastructure providers are embedding SI into both their defensive products and internal development workflows, aiming to reduce exposure to vulnerabilities while maintaining human oversight over final security changes.