# Anthropic’s Claude Mythos Preview model has drawn attention for how quickly it can find software flaws.

*genai · news · 2026-04-26 · Markets Insider*

## Key points

- Anthropic's Claude Mythos model found over 2,000 unknown software vulnerabilities in seven weeks.
- Microsoft and Mozilla directly integrated Mythos findings, leading to record numbers of security fixes.
- U.S. and global financial regulators are urging banks to test Mythos for systemic cybersecurity risks.
- Anthropic's AI agents in simulated markets earned significantly different outcomes based on model strength.
- Users could not detect when weaker AI agents delivered consistently worse deal outcomes in experiments.

**Companies:** Microsoft Corporation, Palo Alto Networks, Anthropic
**Countries:** United States, India, Germany

[Read the full story on Markets Insider](https://markets.businessinsider.com/news/stocks/anthropic-ai-finds-2-000-flaws-as-microsoft-msft-and-palo-alto-networks-panw-race-to-respond-1036063070)

---

Canonical: https://newsio.io/n/516b1564-988b-4599-9f0e-ab2723cd1b3a/anthropics-claude-mythos-preview-model-has-drawn-attention-for-how-quickly-it-ca
Summarized by Newsio from Markets Insider. https://newsio.io/how-it-works