What is Anthropic Project Glasswing?

Anthropic's initiative giving Claude Mythos Preview to 12 partners and 40 open-source groups to scan for zero-days in critical software like OSes and kernels.

Is Claude Mythos Preview public yet?

No — research preview only for now. Hits APIs later at premium pricing; no general release planned soon.

Can Mythos Preview find bugs humans miss?

Yes — nailed 27-year OpenBSD flaw and 16-year FFmpeg bug overlooked by 5 million tests and audits.

🔒 Security & Privacy

Anthropic's Mythos Preview Digs Up a 27-Year OpenBSD Time Bomb

Mythos Preview crushed benchmarks, nailing an OpenBSD flaw from 1999 that dodged 5 million FFmpeg tests. Anthropic's handing it to AWS, Apple, and pals—but is this bug hunter a defender's dream or attacker's blueprint?

theAIcatchup Apr 10, 2026 3 min read

Claude Mythos Preview AI analyzing OpenBSD code for hidden vulnerabilities

⚡ Key Takeaways

Mythos Preview scores 25% better on vuln detection than prior models, autonomously reasoning like experts. 𝕏
Found real zero-days in OpenBSD, FFmpeg, Linux kernel—patched responsibly via partners. 𝕏
Restricted access aims to empower defenders, but dual-use risk looms large in an AI arms race. 𝕏

Published by

theAIcatchup

Community-driven. Code-first.

#AI security #AI security tools #Anthropic #Anthropic Claude #Anthropic Mythos Preview #Claude Mythos Preview #Mythos Preview #Project Glasswing #zero-day vulnerabilities

Worth sharing?

Get the best Open Source stories of the week in your inbox — no noise, no spam.

Originally reported by Dev.to

⚡ Key Takeaways

The 60-Second TL;DR

theAIcatchup

Share this article

Worth sharing?

Related Stories

Project Glasswing: When AI Starts Hunting Its Own Bugs

Anthropic's Mythos AI Cracks Zero-Days Overnight — Locked to Security Elites Only

Anthropic's Claude Mythos: The AI Exploit Machine Locked Away from You

Anthropic's AI Is Flooding Open Source with Real Vulnerability Reports

Stay in the loop