As AI-led attacks multiply, OpenAI launches a new cyber model

OpenAI Expands Daybreak Cyber Defense

In response to a growing number of security incidents involving AI agents, OpenAI is expanding its cyber defense service, Daybreak, with a new tiered structure and a specialized frontier model. The move positions the AI lab as a key vendor in the market for defending against the very technologies it helps create, a trend also seen with competitors like Anthropic and its cyber-focused model, Mythos. This development underscores a critical dynamic in the industry: as AI capabilities advance for both benign and malicious uses, the creators of these models are increasingly bundling them into enterprise security packages.

Blue and Red Tiers for Defensive Operations

The updated Daybreak service now offers two distinct tiers, Blue and Red, both providing approved customers with access to OpenAI's limited-access frontier cyber models. The Blue tier is positioned as the standard offering for most organizations, focusing on core defensive tasks. In contrast, the Red tier provides a more advanced toolkit for proactive security measures, including exclusive access to the new GPT‑5.6‑Cyber model, which is built on the GPT-5.6 Sol architecture.

• Daybreak Blue: The recommended starting point, offering services for incident response, malware analysis, and patch validation.
• Daybreak Red: Provides purpose-trained models for security testing, vulnerability research, and access to the new GPT‑5.6‑Cyber model.

A Market of Self-Made Problems and Solutions

This expansion highlights a broader market trend where AI labs are monetizing solutions to problems their own technology could exacerbate. OpenAI’s own messaging warns that "threat actors will increasingly use AI to conduct cyberattacks," creating a sense of urgency for defenders to adopt similar tools. By offering these capabilities to trusted partners first—reportedly including Accenture, IBM, Crowdstrike, and Cloudflare—OpenAI is building an enterprise ecosystem around its security offerings. While enterprises may see value in sourcing protection from the labs with firsthand knowledge of the risks, critics note the inherent conflict in profiting from both the proliferation and the containment of powerful AI technologies.

The cybersecurity market is becoming a closed loop for major AI labs, which are uniquely positioned to sell high-margin defensive services against threats supercharged by their own foundational models. This creates a powerful, self-reinforcing business model where the pace of model releases directly fuels the demand for corresponding security products.