Anthropic accidentally publishes source code for the AI agent Claude

Anthropic PBC accidentally published the source code for its artificial intelligence agent Claude, raising concerns about its operational security. According to a spokesperson, the incident was caused by human error and did not involve the exposure of sensitive customer data.

This is the second blunder in just a week, after the company made thousands of files available to the public, including details about an internal model called 'Mythos' or 'Capybara', which present cybersecurity risks. The leak of the source code has sparked discussions in the developer community, which is analyzing the implications of this unintended exposure and possible security vulnerabilities.

Anthropic has announced that it will implement measures to prevent the recurrence of these incidents.