After Anthropic launched Claude Code’s 2.1.88 replace, customers rapidly found that it contained a package deal with a supply map file containing its TypeScript codebase, with one particular person on X calling consideration to the leak and posting a file containing the code. The leaked information reportedly comprises greater than 512,000 traces of code and supplies a glance into the inside workings of the AI-powered coding device, as reported earlier by Ars Technica and VentureBeat.
Although Anthropic later fastened the difficulty, that didn’t cease customers from copying the code to a repository on GitHub, which has since amassed greater than 50,000 forks (or copies of the repository). Anthropic launched Claude Code in February of 2025, and the device picked up extra steam after including agentic capabilities that carry out duties on a consumer’s behalf.
“Earlier at present, a Claude Code launch included some inner supply code. No delicate buyer information or credentials have been concerned or uncovered,” Anthropic spokesperson Christopher Nulty says in an emailed assertion to The Verge. “This was a launch packaging problem brought on by human error, not a safety breach. We’re rolling out measures to stop this from taking place once more.”
Arun Chandrasekaran, an AI analyst at Gartner, tells The Verge that whereas the Claude Code leak poses “dangers similar to offering unhealthy actors with doable shops to bypass guardrails,” its long-term impression might be restricted to serving as a “name for motion for Anthropic to speculate extra in processes and instruments for higher operational maturity.”
