## Anthropic's Mythos AI Leaks to Private Forum Before Pentagon Briefing, Raising Security Questions
A small group of unauthorized users gained access to Anthropic PBC's Mythos AI model on the same day the company announced plans for limited testing, according to people familiar with the matter and documentation viewed by Bloomberg News. The model, which Anthropic has described as powerful enough to enable dangerous cyberattacks, was obtained by members of a private online forum who have used it regularly since that initial breach—though not for cybersecurity purposes, the people said.

Rep. Seth Magaziner (R-RI) said Anthropic is taking the right approach by warning the public about what it discovered and by coordinating with the Pentagon. The incident has intensified pressure on Congress to develop a federal framework for AI oversight, particularly around models with potential national security implications. Anthropic has not disclosed the full scope of the breach publicly, but the unauthorized access occurred before the company's formal engagement with defense officials.

The Mythos leak highlights gaps in existing controls for frontier AI models, which companies increasingly restrict before broader release. Industry observers say the episode could accelerate regulatory momentum in Washington, where legislators have struggled to craft rules that balance innovation with security concerns. The case may also influence how other AI developers handle pre-release testing and information-sharing with government agencies.
---
- **Source**: Bloomberg Markets
- **Sector**: The Lab
- **Tags**: Anthropic, Mythos, AI security breach, Pentagon, Congress
- **Credibility**: unverified
- **Published**: 2026-04-22 19:57:30
- **ID**: 76068
- **URL**: https://whisperx.ai/en/intel/76068