What is Claude Mythos Preview?

Claude Mythos Preview is Anthropic's most powerful AI model, announced April 7, 2026. It's a general-purpose model with unprecedented cybersecurity capabilities, scoring 93.9% on SWE-bench Verified and autonomously discovering thousands of zero-day vulnerabilities across every major OS and browser.

Why won't Anthropic release Claude Mythos publicly?

Anthropic deemed Mythos too dangerous for public release due to its cybersecurity capabilities. The model can autonomously find and exploit zero-day vulnerabilities at a rate 181x higher than Opus 4.6. It's the first model since GPT-2 that a lab has restricted from general availability.

What is Project Glasswing?

Project Glasswing is Anthropic's initiative to use Claude Mythos Preview defensively. 40 organizations including Amazon, Apple, Microsoft, CrowdStrike, and the Linux Foundation get early access to find and patch vulnerabilities before attackers can exploit them.

What vulnerabilities did Claude Mythos discover?

Mythos found thousands of zero-days including a 27-year-old OpenBSD TCP SACK denial-of-service bug, a 16-year-old FFmpeg H.264 vulnerability, and a 17-year-old FreeBSD NFS remote code execution flaw (CVE-2026-4747). It found 595 crashes at severity tiers 1-2 and achieved full control flow hijack on 10 separate targets.

How does Claude Mythos compare to Claude Opus 4.6?

On the Firefox 147 benchmark, Mythos developed working exploits 181 times vs just 2 for Opus 4.6. On SWE-bench Verified it scored 93.9% vs Opus 4.6's lower scores. Anthropic described it as a 'step change' in capabilities.

How was Claude Mythos leaked before announcement?

In late March 2026, Anthropic accidentally exposed nearly 3,000 internal documents through a misconfigured content management system, revealing Mythos's existence. Anthropic confirmed it as human error. The official announcement came April 7.

Can I access Claude Mythos Preview?

No. Claude Mythos Preview is restricted to 40 partner organizations under Project Glasswing. Anthropic has stated they do not plan to make it generally available. Access is limited to approved security researchers and critical infrastructure partners.

← Back to dashboard

clawsmith.com/signal/claude-mythos-preview-project-glasswing-zero-days

🔥 HypeUnknownClaude/AnthropicLive

Claude Mythos Preview: Anthropic Reveals Model Too Dangerous to Release, Finds Thousands of Zero-Days Autonomously

Anthropic revealed Claude Mythos Preview on April 7, 2026. A general-purpose model with unprecedented cybersecurity capabilities that autonomously discovered thousands of zero-day vulnerabilities across every major OS and browser, including a 27-year-old OpenBSD bug and a 17-year-old FreeBSD RCE (CVE-2026-4747). 93.9% SWE-bench Verified. Developed working Firefox exploits 181 times vs 2 for Opus 4.6. Restricted to 40 partners under Project Glasswing. First model since GPT-2 deemed too dangerous for public release.

Product Idea from this Signal

A security scanner that uses AI models to discover unknown vulnerabilities in OpenClaw skills and gateway configurations before attackers exploit them

3.8k ▲

OpenClaw has accumulated 138+ CVEs in 63 days, averaging 2.2 new vulnerabilities per day. Anthropic's Claude Mythos Preview found thousands of zero-days across major software, earning 2,342 HN engagement and proving AI-powered vulnerability discovery works at scale. But Mythos is restricted to enterprise partners. 44,000+ ClawHub skills and 500,000+ running OpenClaw instances need proactive security scanning using available AI models. Existing tools only patch known CVEs or detect active compromise. Nothing scans for unknown vulnerabilities in the skill and plugin code before they get exploited.

CLIOPEN-SOURCESECURITYAI-SECURITYDEVTOOL

CompetitiveView Opportunity →