Cyber Risks · The Hacker News

Project Glasswing Proved AI Can Find the Bugs. Who's Going to Fix Them?

Summary: Anthropic postponed the public release of Project Glasswing, an AI model that found vulnerabilities across every major operating system and browser, including one that had sat for 27 years in OpenBSD, and gave access to Apple, Microsoft, Google, Amazon and others to patch bugs before adversaries find them. The article reports that fewer than 1% of the vulnerabilities Mythos found were patched, and that a threat actor using a custom MCP server hosting an LLM compromised 2,516 organizations across 106 countries in parallel, with the attack chain from initial access to exfiltration running autonomously.

What's compelling: Fewer than 1% of the vulnerabilities the most powerful discovery engine ever built surfaced were actually patched, so the finding problem is solved while the fixing problem is not. A single autonomous LLM-driven chain already reached 2,516 organizations in 106 countries, and Mythos-class discovery adds volume that periodic, human-initiated patch cycles were never built to absorb.

Read the original at The Hacker News ↗

Topics
Vulnerabilities & Patching
Related