Cloudflare uses OpenAI's Daybreak to find vulnerabilities
Cloudflare is combining OpenAI models, accessed through Daybreak, with live traffic data from its network to find code vulnerabilities and propose fixes, an early sign of AI security tooling reaching production use.
Cloudflare, which proxies roughly 20% of web traffic, is using OpenAI models accessed through Daybreak in a new early-access service called Vulnerability Discovery and Remediation. OpenAI described the deployment in a customer case study. The service is offered through Cloudflare Managed Defense.
The models examine code that customers have authorized for review, flag potential vulnerabilities and check the evidence behind each finding. Cloudflare then adds context from production: whether the affected route is active, whether it is receiving suspicious traffic, and whether existing security controls already cover it. The service drafts code patches and, where appropriate, custom web application firewall rules to reduce exposure while a fix is pending. Cloudflare runs automated checks on these proposals, and its own staff review them before they go to customers, who decide which changes to apply.
"Traditional security teams have a ton of useful data. They just don't have the people and time to deal with it," said Anderson, a Cloudflare executive. The case study does not give his full name or title.
The case study also covers Cloudflare's internal use of Codex. According to Anderson, about a third of Cloudflare's staff are software developers, and all of them use it. He said a team of four developers used Codex to rebuild a product they had been putting off. The work took three days and cost $1,100 in tokens. He estimated that it would otherwise have taken months and "maybe 10x more developers."
OpenAI did not say when the security service will become generally available or how it is priced. Cloudflare's network blocked an average of 234 billion threats per day in Q1 2026.