OpenAI launches Dawn, an AI initiative centered on discovering and patching vulnerabilities earlier than attackers uncover them. Dawn makes use of the Codex Safety AI agent, launched in March, to create risk fashions based mostly on a corporation’s code, concentrate on attainable assault paths, validate potential vulnerabilities, and automate the detection of high-risk vulnerabilities.
The launch comes simply over a month after rival Anthropic unveiled Claude Mythos, a security-focused AI mannequin. The corporate claimed that the AI mannequin was shared privately solely as a part of its personal initiative, referred to as Mission Glasswing, as a result of it was too harmful to launch publicly. Nonetheless, this didn’t forestall entry by a minimum of a small variety of unauthorized events.
Nonetheless, OpenAI has not had an identical safety product till now. Like Glasswing, Dawn is not constructed on a single AI mannequin. OpenAI says, “Dawn brings collectively probably the most succesful OpenAI fashions, Codex, and safety companions.”
Dawn additionally contains specialised cyber fashions reminiscent of GPT-5.5 and GPT-5.5-Cyber with Trusted Entry for Cyber, which we started rolling out final week. OpenAI additionally mentioned it’s working with “trade and authorities companions” to arrange to “deploy extra cyber-enabled fashions.”
