Anthropic Accidentally Leaked Its Own Claude Code. Now the Company Is Scrambling to Contain the Damage.
The AI company issued over 8,000 copyright takedown requests after accidentally exposing proprietary instructions for how Claude works.
Anthropic just handed its competitors a gift. The AI company accidentally exposed the proprietary instructions behind Claude Code this week when it updated the tool and posted a file to GitHub that linked back to source code that outsiders could download. Within hours, copies were multiplying across the programming platform.
Now the company is racing to stop the bleeding. Anthropic issued over 8,000 copyright takedown requests on Wednesday. The leak didn’t expose customer data, but it handed rivals a detailed roadmap to clone Claude Code’s features without reverse-engineering them. Per Axios, those features include:
- Claude’s ability to review a user’s past session to look for improvements and transfer those learnings across conversations.
- A “persistent assistant” that keeps working in the background even after the user has stopped.
- Remote capabilities, like controlling Claude from a phone or another browser.
The leak is a blow for Anthropic, which recently closed funding at a $380 billion valuation ahead of a possible IPO this year.
Anthropic just handed its competitors a gift. The AI company accidentally exposed the proprietary instructions behind Claude Code this week when it updated the tool and posted a file to GitHub that linked back to source code that outsiders could download. Within hours, copies were multiplying across the programming platform.
Now the company is racing to stop the bleeding. Anthropic issued over 8,000 copyright takedown requests on Wednesday. The leak didn’t expose customer data, but it handed rivals a detailed roadmap to clone Claude Code’s features without reverse-engineering them. Per Axios, those features include:
- Claude’s ability to review a user’s past session to look for improvements and transfer those learnings across conversations.
- A “persistent assistant” that keeps working in the background even after the user has stopped.
- Remote capabilities, like controlling Claude from a phone or another browser.
The leak is a blow for Anthropic, which recently closed funding at a $380 billion valuation ahead of a possible IPO this year.
