OpenAI Launches Daybreak Cybersecurity Platform to Rival Anthropic

A New Era of AI-Powered Cyber Defence

OpenAI on Monday unveiled Daybreak, a cybersecurity platform that combines its most capable AI models with its Codex Security agent framework to help organisations detect, validate, and fix software vulnerabilities at scale. The launch represents OpenAI's most ambitious move yet into the cybersecurity market and positions the company as a direct rival to Anthropic's Project Glasswing initiative.

How Daybreak Works

Daybreak integrates GPT-5.5 and the Codex Security agent, first released in research preview in March under the codename Aardvark, into a unified security operations platform. The system builds an editable threat model from an organisation's code repository, mapping trust boundaries and exposure points to prioritise vulnerabilities based on real-world impact rather than generic heuristics. The platform then validates its findings by executing proof-of-concept exploits within sandboxed environments and generates contextual patches designed to minimise regressions.

The initiative ships with three access tiers: a base tier running standard GPT-5.5 for general security workflows, a Trusted Access for Cyber tier with loosened model safeguards for defensive operations, and a GPT-5.5-Cyber tier for specialised offensive security work such as red teaming, penetration testing, and binary reverse engineering. OpenAI noted that Codex Security has already contributed to fixing over 3,000 critical and high-severity vulnerabilities since its beta period.

A Growing Partner Network

Daybreak is backed by a broad coalition of cybersecurity firms. Eight companies have joined the initiative, including Cloudflare, Cisco, CrowdStrike, Palo Alto Networks, Oracle, and Zscaler. The partner strategy mirrors that of Anthropic's Project Glasswing, announced on 7th April, which recruited AWS, Apple, Google, Microsoft, and Nvidia to use its Claude Mythos Preview model for defensive security work. Anthropic committed up to $100 million in usage credits for that effort.

The Competitive Landscape

The duelling launches underscore a broader race among AI labs to dominate the cybersecurity market. UK AI Safety Institute evaluations found GPT-5.5 and Anthropic's Mythos Preview perform at comparable levels on cyber tasks. Palo Alto Networks warned in a recent blog post that frontier AI models have compressed the time from initial access to data exfiltration to as little as 25 minutes, making AI-assisted defence a necessity rather than an option. OpenAI requires all approved users and organisations to implement phishing-resistant authentication by 1 June 2026.