Anthropic’s Claude Mythos Finds Thousands of Zero-Day Flaws Across Major Systems
Anthropic’s Claude Mythos Finds Thousands of Zero-Day Flaws Across Major Systems
https://thehackernews.com/2026/04/anthropics-claude-mythos-finds.html
Publish Date: 2026-04-08 05:16:00
Source Domain: thehackernews.com
- Anthropic introduces Project Glasswing to use its frontier model, Claude Mythos, to identify and remedy security vulnerabilities in collaboration with major tech firms.
- The model, which has found thousands of high-severity zero-day vulnerabilities, demonstrates capabilities that can surpass skilled human programmers in exploiting software vulnerabilities.
- Due to potential risks and capabilities of misuse, Anthropic has opted not to make the model publicly accessible.
- Project Glasswing aims to harness the model’s capabilities for defensive purposes before they can be used maliciously, committing up to $100 million in usage credits for Mythos Preview.
- Anthropic emphasizes that the model’s dangerous capabilities emerged from general code, reasoning, and autonomy improvements, making it dual-purpose in its effects.
- Recent leaks have exposed vulnerabilities in Anthropic’s other products, including a bypass of safety measures in its AI coding agent, raising further concerns about security handling.
