Anthropic confirmed it is testing a new AI model called Claude Mythos. The company describes the performance of the model as a step change.
Internal documents were exposed in a public data store due to a content management system configuration error. Anthropic attributed the disclosure to human error.
Leaked files identify Mythos as part of a new Capybara tier. The model scores significantly higher than the current Opus model in coding, reasoning, and cybersecurity.
Anthropic is limiting initial testing to a small group of early access customers. The company is prioritizing cyber defense organizations to manage risks associated with the model's advanced capabilities.