Anthropic Investigates Unauthorized Access to Its New AI Model 'Mythos'
Anthropic is investigating reports that an unauthorized group accessed its advanced security-focused AI model 'Mythos'. The news has prompted Japanese officials to hold urgent talks with major financial institutions.
📋 Article Processing Timeline
- 📰 Published: April 22, 2026 at 15:34
- 🔍 Collected: April 22, 2026 at 16:02 (27 min after Published)
- 🤖 AI Analyzed: April 22, 2026 at 19:47 (3h 45m after Collected)
Reuters, citing Bloomberg, reported that a small group of unauthorized users obtained access to Anthropic's cutting-edge AI model 'Mythos'. The breach occurred on the same day Anthropic announced a limited preview for corporate testing. Sources indicate the users have been utilizing Mythos for purposes other than cyber defense. An Anthropic spokesperson confirmed the investigation, stating they are looking into unauthorized access via a third-party vendor environment. Mythos is part of 'Project Glasswing,' a controlled initiative intended to allow specific institutions to use the model for defensive cybersecurity. The model's unprecedented ability to identify security vulnerabilities has raised significant regulatory concerns. In related news, Japanese Finance Minister Satsuki Katayama plans to meet with representatives from Mitsubishi UFJ, Sumitomo Mitsui, and Mizuho financial groups as early as this week to discuss the Mythos model and its potential impact on the banking sector.