OpenAI launches Daybreak, the AI ​​shield that repairs the web

This move positions OpenAI in direct competition with Anthropic, which just over a month ago had shaken up the sector by introducing Claude Mythos within the Project Glasswing initiative. Although Anthropic initially restricted access to its model for security reasons, considering it too sensitive for large-scale public release, some breaches still exposed the technology to unauthorized parties. Daybreak therefore represents the necessary answer to fill a gap in OpenAI’s offering, proposing an approach that is not limited to a single tool, but coordinates a plurality of advanced resources.

The structure of the initiative brings together various proprietary technologies and external expertise, combining the most capable models of the Codex family with the support of partners specialized in cybersecurity. The operational core makes use of the new vertical models GPT-5.5-Cyber ​​and GPT-5.5 with Trusted Access for Cyber, whose controlled distribution officially began last week. The active involvement of industrial and government partners confirms OpenAI’s desire to consolidate its role in the defense of critical infrastructures, preparing the ground for the deployment of systems equipped with increasingly sophisticated IT capabilities.