# Anthropic, a software-focused company, says its new AI model is "strikingly capable at computer security tasks" Anthropic’s testers were.

*fintech · news · 2026-04-08 · New York Magazine*

## Key points

- Anthropic's new Claude Mythos Preview model can identify and exploit zero-day vulnerabilities in major software.
- Non-experts using Claude Mythos Preview were able to find and exploit sophisticated security vulnerabilities.
- Anthropic has launched Project Glasswing to use its model for securing critical software and industry preparation.
- Anthropic is donating $100M in usage credits to defenders to help patch vulnerabilities before public release.
- The model's capabilities create a temporary power imbalance, giving Anthropic exclusive offensive and defensive cyber abilities.

**Companies:** Anthropic
**Countries:** United States

[Read the full story on New York Magazine](https://nymag.com/intelligencer/article/anthropic-claude-mythos-preview-cybersecurity.html)

---

Canonical: https://newsio.io/n/d2eeddca-0402-4eca-9e15-7205b5efc387/anthropic-a-software-focused-company-says-its-new-ai-model-is-strikingly-capable
Summarized by Newsio from New York Magazine. https://newsio.io/how-it-works