Anthropic Investigates Unauthorized Mythos Access

The security breach has raised concerns among global policymakers about the safety of frontier AI models.

While Anthropic claims there is no evidence its systems were compromised, documentation viewed by researchers suggests the model has been accessed regularly.

Mythos is specifically designed to identify and exploit software vulnerabilities, making its potential misuse a significant national security risk.

Central banks and cybersecurity agencies in several countries are now monitoring the situation closely.