SAN FRANCISCO ? Anthropic spent the end of March and the first days of April trying to contain an accidental Claude Code source leak that spilled onto GitHub and then triggered a much broader takedown than the company said it intended.

In TechCrunch's April 1 report, Anthropic said a copyright notice aimed at leaked code ended up affecting more than 8,000 repositories before the company retracted most of the request and GitHub restored access to many of the affected forks.

The takedown followed a separate internal disclosure problem. TechCrunch's March 31 account said Anthropic had accidentally exposed nearly 3,000 internal files and a large amount of Claude Code source material in a recent software release, creating a second public misstep in less than a week.

The cleanup came as Anthropic was already tightening the business model around Claude Code. In a later report, TechCrunch wrote that Anthropic would start charging separately for third-party harness usage such as OpenClaw, a move company executives said reflected engineering and growth constraints.

Taken together, the leak, the takedown backlash and the pricing change marked a difficult stretch for one of the companies most actively selling AI coding tools to professional developers.