OpenAI has unveiled Daybreak, an innovative artificial intelligence initiative aimed at identifying and resolving security vulnerabilities before they can be exploited by cyber attackers. This initiative leverages the capabilities of the Codex Security AI agent, introduced earlier in March, to generate a threat model based on a company’s code. Daybreak concentrates on potential attack vectors, verifies likely vulnerabilities, and automates the detection of those posing the highest risk.
This launch follows closely on the heels of Anthropic’s announcement of Claude Mythos, a security-focused AI model. Anthropic labeled this model as too hazardous for public release, opting instead for a controlled distribution under their Project Glasswing initiative. Despite these precautions, unauthorized access by several entities was reported.
Until now, OpenAI had not offered a comparable security solution. Daybreak, much like Glasswing, is not reliant on a single AI model. According to OpenAI, “Daybreak integrates the most advanced OpenAI models, Codex, and collaborates with our security partners” to deliver robust security solutions.
The Daybreak initiative also includes specialized cyber models such as GPT-5.5 with Trusted Access for Cyber and GPT-5.5-Cyber, which began their rollout last week. OpenAI is actively engaging with both industry and government partners as it prepares for the deployment of increasingly sophisticated cyber-capable models.
