OpenAI takes on Anthropic’s Claude Mythos with new cybersecurity AI system ‘Daybreak’

After Artificial Intelligence (AI)-powered enterprise agents, the race has shifted towards AI cybersecurity models with tech giants and AI startups bringing security-focused AI systems to detect threats, strengthen digital infrastructure, and defend against cyberattacks. After Anthropic’s Claude Mythos, OpenAI has released “Daybreak” as a new “frontier AI for cyber defenders,” helping organisations build safer software.

AI models, like the Claude Mythos, have created global cybersecurity concerns around the misuse of their capabilities. As the government creates a plan to strengthen its cyber defences, OpenAI’s Daybreak is expected to emerge as the company’s response in the growing AI cybersecurity race.

Must read: Anthropic’s Claude Mythos triggers call for sovereign AI infrastructure in India: All details

OpenAI said the era of cyber defence should not only be about finding and patching vulnerabilities. However, future software systems should be built in a way that they can continue functioning safely and remain secure even if some vulnerabilities still exist.

OpenAI’s Daybreak features and capabilities

OpenAI, in a blog post, revealed that Daybreak combines its AI models, Codex’s agentic capabilities, along with industry partnerships to support AI-powered cyber defence. The AI platform is said to offer tools for secure code review, threat modelling, vulnerability detection, patch validation, malware analysis, remediation guidance, and automated monitoring and response systems.

In addition, Daybreak can help organisations identify high-risk vulnerabilities faster, generate and test patches directly within repositories, verify fixes with audit-ready validation systems, and automate threat detection workflows. 

Must read: US govt could review AI models before public release amid Mythos concerns

Daybreak can be integrated into existing development environments, empowering software systems to be more resilient throughout the development cycle. OpenAI is also working with government and industry partners to safely deploy advanced cyber-capable AI models in the coming weeks.

Daybreak will work under three models: GPT-5.5, GPT-5.5 with Trusted Access for Cyber, and GPT-5.5-Cyber. The GPT 5.5 will remain the default model for general-purpose and standard safeguards. The GPT-5.5 with Trusted Access for Cyber will be used for defensive security workflows, including secure code review, vulnerability triage, malware analysis, detection engineering, and patch validation. Lastly, the new GPT-5.5-Cyber will be the most advanced option for specialised authorised workflows.