OpenAI has simply launched Daybreak, a cybersecurity initiative that is clearly the corporate’s competitor to Anthropic’s Project Glasswing. Should you’ll recall, Glasswing makes use of Anthropic’s unreleased AI mannequin, Claude Mythos Preview, to offer its purchasers’ cyber protection wants. It has been promising, thus far: Mozilla revealed in April that Mythos helped it find and patch 271 vulnerabilities within the newest launch of the Firefox browser. OpenAI says Dawn makes use of its varied AI fashions, together with its specialised safety agent Codex.
In its announcement, the corporate defined that Dawn is constructed across the premise that cyber protection ought to be constructed into software program from the beginning and never simply revolve round discovering and fixing vulnerabilities. Dawn goals to prioritize high-impact points and scale back hours of research to minutes, to generate and take a look at patches inside repositories and to ship again outcomes with audit-ready proof to the purchasers’ programs. In OpenAI’s instance, it requested Codex Safety to scan a codebase, validate the highest-risk findings and repair them.
Discover and repair vulnerabilities earlier with Dawn pic.twitter.com/yobOSWYeWP
— OpenAI (@OpenAI) May 11, 2026
Dawn will use GPT-5.5 for normal functions and GPT-5.5 with Trusted Entry for Cyber for many defensive safety workflows, together with “safe code evaluation, vulnerability triage, malware evaluation, detection engineering and patch validation.” It is going to additionally depend on GPT-5.5-Cyber for “preview entry for specialised workflows, together with licensed crimson teaming, penetration testing and managed validation.” OpenAI is already working with a number of companions underneath the initiative, together with Cloudflare, Cisco, CloudStrike, Palo Alto Networks, Oracle and Akamai.
