--- title: OpenAI launches Daybreak, an AI tool for cyber defenders description: Anthropic's Project Glasswing (Claude Mythos Preview) and OpenAI's Daybreak (GPT-5.5) represent a new class of AI-powered vulnerability discovery platforms.... canonical: https://daily.dev/posts/openai-launches-daybreak-an-ai-tool-for-cyber-defenders-mya66olag twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: OpenAI launches Daybreak, an AI tool for cyber defenders | daily.dev og:description: Anthropic's Project Glasswing (Claude Mythos Preview) and OpenAI's Daybreak (GPT-5.5) represent a new class of AI-powered vulnerability discovery platforms.... og:url: https://daily.dev/posts/openai-launches-daybreak-an-ai-tool-for-cyber-defenders-mya66olag og:image: https://api.daily.dev/og/posts/mya66OlAg.png og:image:alt: OpenAI launches Daybreak, an AI tool for cyber defenders og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # OpenAI launches Daybreak, an AI tool for cyber defenders **[Collections](https://daily.dev/sources/collections)** · 5 min read · 1 upvotes · 0 comments ## Summary Anthropic's Project Glasswing (Claude Mythos Preview) and OpenAI's Daybreak (GPT-5.5) represent a new class of AI-powered vulnerability discovery platforms. Glasswing found over 10,000 vulnerability candidates across 1,000+ open-source projects in its first month, with 1,094 confirmed critical or high severity — but only 97 have been patched. The core problem is a widening gap between AI-accelerated discovery and the ecosystem's capacity to remediate. Both platforms demonstrate exploit chain construction and automated proof-of-concept generation, capabilities that are equally useful to attackers. Access models differ: Glasswing is a restricted consortium while Daybreak uses tiered trust tiers. Practical advice for security teams includes auditing dependency inventories, reducing mean time to remediation, investing in defensive monitoring, and funding open-source maintainers whose underfunded projects represent systemic downstream risk. ## Content ## The race nobody asked for Two of the biggest AI labs have now positioned frontier models as cybersecurity infrastructure, and the results are genuinely impressive and genuinely unsettling at the same time. Anthropic's Project Glasswing, running on an unreleased model called Claude Mythos Preview, found over 10,000 high- or critical-severity vulnerability candidates across 1,000+ open-source projects in its first month. Of 1,726 independently validated findings, 1,094 are confirmed critical or high severity. The most significant is CVE-2026-5194 in WolfSSL — a certificate forgery flaw with a CVSS score of 9.1 affecting IoT, automotive, and industrial systems. Six weeks later, OpenAI launched Daybreak, built on GPT-5.5 and Codex Security, with a similar pitch: automated vulnerability detection, patch validation, and secure code review for enterprises and governments. UK AI Security Institute benchmarks put the two platforms at 71.4% vs 68.6% on expert CTF tasks. Effectively identical. ## What these models can actually do The capability that matters most isn't finding individual bugs — it's chaining them. Cloudflare, one of roughly 50 vetted Glasswing partners, tested Mythos Preview against 50+ of their own repositories and found it particularly good at two things: - **Exploit chain construction**: taking multiple low-severity bugs and combining them into a working exploit - **Automated proof generation**: writing, compiling, and running proof-of-concept code in a loop until something works This is meaningfully different from previous generations of static analysis. Cloudflare found 2,000 bugs with a false positive rate better than human testers. Mozilla found 271 vulnerabilities in Firefox 150 — ten times more than with an earlier Claude model. The UK AISI has been tracking how quickly AI models can complete autonomous, multi-stage penetration tests. In November 2025, the difficulty of tasks AI could handle was doubling every 8 months. By February 2026, that interval had shrunk to 4.7 months. Mythos Preview and GPT-5.5 show even steeper improvement curves. ## The patch gap Here's the problem nobody has solved: of those 1,094 confirmed critical or high-severity vulnerabilities Glasswing found, only 97 have been patched. AI can find vulnerabilities far faster than the ecosystem can fix them. Open-source maintainers — many of them underfunded and working in their spare time — are already overwhelmed by normal bug reports. Flooding them with AI-generated CVEs doesn't help if there's no corresponding investment in remediation capacity. Cloudflare made a point worth sitting with: the industry's focus on speed alone misses something. Patching faster without fixing underlying architecture — defense-in-depth, blast radius reduction, the ability to roll out fixes instantly — just accelerates a treadmill. The same exploit-chaining capability that helps defenders will be used against every application on the internet. Daniel Stenberg, who maintains curl, found that well-maintained codebases with disciplined dependency management proved largely resistant to Mythos-class scanning. That's not a coincidence. It's an argument for funding maintainers, not just buying AI tools. ## The access question Both platforms restrict access, but differently. Glasswing is a walled consortium of about 52 vetted organizations. Mythos Preview remains unreleased publicly — Anthropic says it will stay that way until stronger safeguards exist. Daybreak uses a tiered trust framework: a general-purpose GPT-5.5 variant, a "Trusted Access for Cyber" tier for verified defenders, and a more permissive GPT-5.5-Cyber for red teaming. OpenAI is also in discussions with the European Commission about providing model access for vulnerability identification, which positions it more favorably with regulators than Anthropic's more restricted approach. Cisco, CrowdStrike, and Palo Alto Networks have signed on to both platforms — a deliberate dual-stack strategy rather than picking a winner. When model capability is this close, the access model, the harness architecture, and the partner network become the actual differentiators. ## What this means in practice Bruce Schneier's read is worth noting: the announcement around Mythos is partly marketing. Comparable capabilities exist in GPT-5.5 and smaller models, and cost constraints likely drove some of the access restrictions. The real concern is the broader trend — AI systems improving at finding and exploiting vulnerabilities creates a window where offense outpaces defense, and that window is getting shorter. Google's Threat Intelligence Group recently disclosed the first documented case of a criminal actor using AI to discover and weaponize a zero-day exploit. That's not a hypothetical anymore. Most attacks still exploit known vulnerabilities, credential theft, and identity abuse rather than zero-days. AI raises the baseline for software quality but doesn't eliminate existing exposure or address attack paths that bypass code entirely. The limiting factor for most organizations isn't vulnerability discovery — it's the ability to detect, prioritize, and respond to risk consistently. ## Practical takeaways For security teams, a few things follow from all this: - Audit your dependency inventory against CVE databases now, before someone else does it for you - Measure and reduce mean time to remediation — finding bugs faster only helps if you can fix them faster - Invest in defensive monitoring and blast radius reduction, not just scanning tools - Fund the open-source maintainers whose underfunded projects represent systemic risk for everyone downstream The tools are impressive. The gap between discovery and remediation is the actual problem. ## Similar posts on daily.dev - [Project Glasswing has uncovered 10,000 vulnerabilities: Anthropic](https://daily.dev/posts/project-glasswing-has-uncovered-10-000-vulnerabilities-anthropic-33ik2enrf) · CSO Online · 1 upvotes · 0 comments - [AI just found 10,000 bugs in a month. Finding them is the easy part](https://daily.dev/posts/ai-just-found-10-000-bugs-in-a-month-finding-them-is-the-easy-part-fjluqyxbl) · Swimm · 0 upvotes · 0 comments --- Tags: [#cyber](https://daily.dev/tags/cyber), [#openai](https://daily.dev/tags/openai), [#ai-security](https://daily.dev/tags/ai-security), [#codex](https://daily.dev/tags/codex) [View this post on daily.dev](https://daily.dev/posts/openai-launches-daybreak-an-ai-tool-for-cyber-defenders-mya66olag) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"OpenAI launches Daybreak, an AI tool for cyber defenders","url":"https://daily.dev/posts/openai-launches-daybreak-an-ai-tool-for-cyber-defenders-mya66olag","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/openai-launches-daybreak-an-ai-tool-for-cyber-defenders-mya66olag"},"datePublished":"2026-05-11T21:22:51.694Z","dateModified":"2026-05-24T13:45:16.119Z","description":"Anthropic's Project Glasswing (Claude Mythos Preview) and OpenAI's Daybreak (GPT-5.5) represent a new class of AI-powered vulnerability discovery platforms....","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/openai-launches-daybreak-an-ai-tool-for-cyber-defenders-mya66olag","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":1},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"cyber,openai,ai-security,codex","timeRequired":"PT5M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"OpenAI launches Daybreak, an AI tool for cyber defenders"}]} ```