# US firm Anthropic has claimed its latest model, Claude Mythos, is too dangerous to release due to its exceptional cyber hacking capabilities.

*fintech · news · 2026-04-22 · ABC (Australian Broadcasting Corporation)*

## Key points

- Anthropic's Claude Mythos AI model is withheld from wide release due to extreme cyber attack capabilities.
- Only large US tech companies and about 40 critical infrastructure organizations have access to Mythos via Project Glasswing.
- Critical infrastructure sectors outside the US, like Australian utilities and banks, lack direct access to Mythos for defense.
- AI cybersecurity technology cycles are accelerating, with capabilities improving every three to six weeks.
- Australian regulators and the government are monitoring AI risks but face challenges without direct access to advanced models.

**Companies:** Apple, Microsoft, Amazon Web Services, Cisco
**Countries:** Australia, United States

[Read the full story on ABC (Australian Broadcasting Corporation)](https://www.abc.net.au/news/2026-04-23/powerful-ai-tools-posing-cybersecurity-risks-australia-lagging/106584436)

---

Canonical: https://newsio.io/n/947d43c6-e713-487a-a6e8-91543841f7ee/us-firm-anthropic-has-claimed-its-latest-model-claude-mythos-is-too-dangerous-to
Summarized by Newsio from ABC (Australian Broadcasting Corporation). https://newsio.io/how-it-works