OpenAI introduces Daybreak cyber platform, takes on Anthropic Mythos

OpenAI has unveiled Daybreak, its answer to Anthropic’s Claude Mythos, amid a growing market for frontier AI-powered cyber defense platforms. The initiative combines OpenAI’s large language models, Codex’s agentic capabilities, and integrations with the broader enterprise security ecosystem.

The company said Daybreak is focused on accelerating cyber defense operations and enabling organizations to secure software across the development lifecycle continuously.

Announcing the initiative on X, Sam Altman, CEO at OpenAI, said, “OpenAI is launching Daybreak, our effort to accelerate cyber defense and continuously secure software. AI is already good and about to get super good at cybersecurity; we’d like to start working with as many companies as possible now to help them continuously secure themselves.”

Daybreak takes on Mythos

The surge in AI-driven cyber threats has recently shifted the AI race toward AI cybersecurity models. In April this year, Anthropic unveiled Project Glasswing, built around Claude Mythos Preview. Anthropic described it as a cybersecurity-focused AI system capable of autonomously identifying software vulnerabilities at scale.

While introducing Daybreak, OpenAI explained that deploying AI in modern cyber defense involves three core stages. The first is prioritizing high-impact threats and reducing hours of security analysis to minutes through more efficient AI reasoning and token usage. The second involves generating and testing patches directly within enterprise repositories using scoped access, monitoring, and review. The final stage focuses on sending results and audit-ready evidence back into enterprise systems to track, validate, and verify remediation efforts.

In Daybreak, Codex security is designed to identify and fix vulnerabilities by building an editable threat model from the enterprise’s repository and focusing analysis on realistic attack paths and high-impact code. The system would then validate likely vulnerabilities in an isolated environment. This would help teams to prioritise real, reproducible issues over noisy alerts. This will be followed by automated detection and response, where AI will be able to spot higher-risk vulnerabilities and enable end-to-end automated monitoring.

“The divergence reflects fundamentally different approaches to security and commercialization. OpenAI is positioning Daybreak and GPT-5.5-Cyber as a controlled cyber-defense platform for vetted defenders, focused on operational workflows such as vulnerability detection, patch validation, malware analysis, and secure software development,” said Pareekh Jain, CEO at EIIRTrend & Pareekh Consulting. “Strategically, Daybreak helps OpenAI counter the perception that Anthropic leads in frontier cyber AI. Instead of relying on a single secretive model, OpenAI is building a scalable cyber-defense ecosystem integrated into enterprise workflows and developer environments.”

Jain said Anthropic, by contrast, treats Mythos as a far more sensitive dual-use cyber-intelligence system with stronger offensive reasoning capabilities and higher misuse risks. As a result, access remains tightly restricted to a small set of organizations, influenced both by safety concerns and broader US national-security considerations.

OpenAI’s cybersecurity model stack

OpenAI is pursuing a scalable cyber defense platform strategy with Daybreak and is rolling out the initiative through three different model tiers: GPT-5.5 (default), GPT-5.5 with Trusted Access for Cyber, and GPT-5.5-Cyber.

The standard GPT-5.5 model is positioned for general-purpose enterprise use cases, including developer assistance and knowledge work. GPT-5.5 with Trusted Access for Cyber is designed for defensive security workflows such as secure code review, vulnerability triage, malware analysis, detection engineering, and patch validation.

At the highest tier, GPT-5.5-Cyber will provide preview access for specialized cybersecurity workflows, including authorized red teaming, penetration testing, and controlled validation.

Governments and industry join in

OpenAI said it plans to build Daybreak alongside both industry and government partners as it expands the platform’s cybersecurity capabilities and enterprise reach.

To begin with, Daybreak is being developed alongside partners including Cisco, Oracle, CrowdStrike, Palo Alto Networks, Cloudflare, Fortinet, Akamai, and Zscaler.

At the government level, the European Commission is currently in discussions with OpenAI regarding access to its advanced AI models for identifying cybersecurity vulnerabilities. According to Commission spokesperson Thomas Regnier, OpenAI proactively approached the EU, and discussions are underway around potential next steps, including possible access to the company’s new model. Discussions with Anthropic are also continuing. However, they have not yet reached the same stage as those with OpenAI.

While answering questions during the Commission’s daily press briefing, spokesperson Regnier said the European Commission welcomes OpenAI’s transparency and their intent to give the Commission access to its new model. This will allow the Commission to follow the deployment of this model very closely and also to potentially address certain security concerns in a closer way.

Amit Jaju, senior managing director at Ankura Consulting, said, “OpenAI is actively leveraging its trusted access framework to rapidly build goodwill with European regulators and demonstrate transparency. By offering early access, OpenAI aligns itself closely with upcoming regulatory demands and secures a strategic market position.”  Jaju noted that Anthropic is taking a highly restricted approach, initially sharing its Mythos model only with select US technology partners to patch vulnerabilities first. “Anthropic recognizes the severe risks associated with autonomous AI agents and the potential for the model to be misused to target critical software, choosing to prioritize closed testing over rapid geopolitical expansion.”

The article originally appeared on CSO.