Cloudflare has begun providing early access to "Vulnerability Discovery and Response," a new service designed to detect software vulnerabilities in customer applications and block attacks at the network edge. Integrated into Cloudflare's "Managed Defense" security monitoring service, the tool utilizes OpenAI's cybersecurity-specific model, "GPT-5.6-Cyber," via the Daybreak Defense Network.
The service identifies vulnerabilities by analyzing public paths and security events based on web asset inventory and WAF data, coordinating reconnaissance agents with detection agents. Once a vulnerability is detected, it undergoes verification and is assigned a risk level. Cloudflare stated that human approval is required for applying firewall rules or code patches, and the model will not apply them automatically.
Source: Cloudflare、OpenAIモデル活用の脆弱性検出サービスを早期提供 - 디지털투데이 (Google News: OpenAI)