Cloudflare, which proxies roughly 20% of the web's traffic, is combining OpenAI models accessed through Daybreak with its global network and threat context. The combination powers Vulnerability Discovery and Remediation, an early-access service offered through Cloudflare Managed Defense. The models investigate customer-authorized code, assess potential vulnerabilities and check the evidence behind each finding.
Cloudflare adds live traffic data, security signals and validation to prioritize findings. The service then prepares proposed code patches and, where appropriate, custom web application firewall rules to reduce exposure while developers review a fix. Cloudflare runs automated checks on these proposals, and its team reviews them before sharing them with customers, who retain control over which changes to implement.
Cloudflare also uses Codex across its engineering organization. Mark Anderson, President of Revenue at Cloudflare, says a third of the company are software developers and 100% of them use Codex to boost output and deliver better code. He describes a rebuild the team had postponed for lack of manpower that a four-person team completed in three days for $1,100 in tokens, work he estimated would otherwise have taken months and "maybe 10x more developers."
Anderson argues that the automation lets machines handle routine work so humans can focus on the most important engineering and security decisions. For Cloudflare customers, the result is a clearer path from identifying a vulnerability to protecting the affected application.
Source: OpenAI Blog · Summarized by HeadlinesBriefing