OpenAI has expanded Daybreak, its cyber-defence service, launching a new AI model built specifically to help organisations defend against increasingly common AI-driven cyberattacks. The move comes amid a wave of incidents involving AI agents behaving maliciously, from compromising platforms to socially engineering intrusions, and follows Anthropic's earlier release of its own cyber-focused model, Mythos. OpenAI argues that as attackers increasingly use AI to launch faster, more autonomous attacks, defenders face a "narrowing window" to prepare, positioning the update as a necessary response to a fast-changing threat landscape.

Daybreak now offers two tiers, Blue and Red, both granting approved customers access to OpenAI's restricted frontier cyber models. Blue, described as the recommended starting point for most defenders, covers incident response, malware analysis and patch validation, while Red provides a broader toolkit for security testing and vulnerability research, including the new GPT‑5.6‑Cyber model built on GPT‑5.6 Sol. GPT‑5.6‑Cyber is currently limited to "trusted customer partners" such as Accenture, IBM, CrowdStrike and Cloudflare. Critics note that such offerings also serve as marketing for AI labs, even as enterprises increasingly turn to those same companies for protection given their first-hand knowledge of AI-related security risks.