On April 7, 2026, Anthropic publicly disclosed the Claude Mythos Preview — a model it describes as the single largest step in frontier AI capabilities. During internal testing, the system autonomously identified thousands of high-severity zero-day vulnerabilities across every major operating system and web browser, prompting the company to launch Project Glasswing, a defensive cybersecurity coalition backed by up to $100 million in Mythos usage credits for defenders.
Cybergym Score of 83.1% Leaves Previous Models in the Dust
Benchmark results show a stark gap. Claude Mythos Preview hit 83.1% on Cybergym versus 66.6% for Claude Opus 4.6. On SWE-bench Verified it scored 93.9% against 80.8%, and on SWE-bench Pro it achieved 77.8% compared to 53.4% — a 24-point spread. On Humanity's Last Exam without tools, Mythos registered 56.8% while its predecessor managed 40.0%. Anthropic stresses these gains come from broader advances in reasoning, multi-step planning, and autonomous agentic behavior, not cybersecurity-specific training.
Given a target codebase in an isolated container, the model reads source code, forms hypotheses about memory-safety flaws, compiles and runs the software, employs debuggers like Address Sanitizer, ranks files by vulnerability likelihood, and produces validated bug reports with working proof-of-concept exploits — many with minimal human direction.
27-Year-Old OpenBSD Bug and 16-Year-Old FFmpeg Flaw Found in Hours
According to Tom's Hardware, a 27-year-old OpenBSD TCP SACK vulnerability — a subtle integer overflow enabling remote crash of any responding host — was found autonomously after roughly 1,000 runs at a total cost under $20,000. A 16-year-old FFmpeg H.264 bug survived more than five million automated tests and multiple audits before Mythos caught it.
Browser results drew particular attention. On Firefox 147 JavaScript engine testing, Mythos produced 181 full shell exploits and 29 register-control cases; Claude Opus 4.6 produced two shell exploits across the same test set. The model also built working Linux kernel privilege-escalation chains (user to root) after filtering 100 recent CVEs down to 40 exploitable candidates and successfully exploiting more than half.
Human validators reviewed 198 of the model's vulnerability reports and agreed with its severity ratings 89% of the time, with 98% agreement within one severity level.
Project Glasswing: $100M AI Credits and 11 Founding Partners
Fewer than 1% of the identified bugs have been fully patched. Anthropic is coordinating responsible disclosure with cryptographic SHA-3 commitments for unpatched issues, following a 90-plus-45-day timeline before releasing full details. Named examples already in disclosure include CVE-2026-4747, a 17-year-old FreeBSD NFS server remote code execution bug granting unauthenticated root access.
Project Glasswing, announced alongside the model, counts founding partners including Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, Nvidia, and Palo Alto Networks. Access is being extended to more than 40 additional critical software organizations. Anthropic also committed $4 million in open-source security donations: $2.5 million to Alpha-Omega via the OpenSSF through the Linux Foundation, and $1.5 million to the Apache Software Foundation.
The company acknowledged that AI tools like Mythos lower the barrier for finding and exploiting vulnerabilities, flagging near-term risk from state actors (China, Iran, North Korea, Russia) and criminal groups if similar capabilities spread without controls. It described a period of transitional turmoil before defenders fully integrate the technology. Anthropic said upcoming Claude Opus releases will include safeguards to detect and block dangerous cybersecurity outputs, and plans to introduce a Cyber Verification Program for vetted security professionals. A public report on partner findings and patched vulnerabilities is expected within 90 days.