Subsequent to Anthropic’s deployment of Claude Code’s 2.1.88 update, it was rapidly uncovered by users that a package within it harbored a source map file showcasing its TypeScript codebase. An individual on X then drew attention to this disclosure by sharing a file containing the code. Reportedly, the divulged data encompasses upwards of 512,000 code lines, offering a glimpse into the operational core of the AI-enhanced coding instrument, a detail previously brought to light by Ars Technica and VentureBeat.
While Anthropic later resolved the anomaly, it did not prevent individuals from duplicating the source into a GitHub repository, a repository that has since gathered over 50,000 forks (or replications). The introduction of Claude Code by Anthropic occurred in February 2025, with the instrument gaining considerable traction after integrating agentic features designed to carry out duties for its users.
“This morning, a Claude Code update inadvertently comprised some internal source code. Sensitive customer details or login credentials were neither compromised nor disclosed,” Christopher Nulty, a spokesperson for Anthropic, communicated in an email statement to The Verge. “The incident stemmed from a packaging error in the release, attributed to human mistake, not a breach of security. Steps are being enacted to preclude a similar occurrence in the future.”
According to Arun Chandrasekaran, an AI analyst at Gartner, who spoke to The Verge, the divulgence of Claude Code, while introducing “dangers such as furnishing nefarious entities with potential means to circumnavigate protective measures,” its lasting repercussions may merely amount to an “exhortation for Anthropic to allocate greater resources toward procedures and utilities for superior operational refinement.”
{content}
Source: {feed_title}