# US, UK and Canadian governments are raising concerns over the model’s abilities.

*genai · news · 2026-04-13 · Siliconrepublic.com*

## Key points

- Anthropic's Claude Mythos Preview is significantly better at generating exploits than previous models.
- Mythos was released only to about 40 key tech and finance companies, not the public.
- Government agencies in the US, UK, and Canada have explicitly warned banks about Mythos's cybersecurity risks.
- Mythos created working exploits 181 times, while Opus 4.6 almost never succeeded.
- Major entities like JP Morgan, Goldman Sachs, and Google are actively participating in testing Mythos under project Glasswing.

**Companies:** JP Morgan Chase, Goldman Sachs, Citigroup, Bank of America, Morgan Stanley, Amazon Web Services, Apple, Google, Microsoft, Nvidia, Anthropic, CrowdStrike, Palo Alto Networks, Broadcom, Cisco
**Countries:** US, UK, Canada

[Read the full story on Siliconrepublic.com](https://www.siliconrepublic.com/enterprise/anthropic-claude-mythos-preview-us-canada-uk-banks-cybersecurity)

---

Canonical: https://newsio.io/n/e7f9d71a-48c4-42bc-a206-0cc11ed7cc47/us-uk-and-canadian-governments-are-raising-concerns-over-the-models-abilities-jp
Summarized by Newsio from Siliconrepublic.com. https://newsio.io/how-it-works