AI Threat Watch

Case claude-code-leak-luresModel & data leakWorldwide

Claude Code source map leak followed by malware-laced fake 'leak' repos

Anthropic accidentally shipped a source map exposing ~513,000 lines of Claude Code; attackers then used fake leaked-code GitHub repos to spread malware to developers.

HighAttack on AIStatus: ContainedVerified case file

Open in the interactive mapCite this case

Reported
Mar 31, 2026
Location
Worldwide
Target
Anthropic Claude Code / developers
Severity
High
Triggered by
Anthropic packaging error; GitHub user idbzoomh spreading Vidar and GhostSocks
Actor type
Criminal
Reported loss
None reported
CVE
None

What happened

Anthropic accidentally shipped a source map exposing ~513,000 lines of Claude Code; attackers then used fake leaked-code GitHub repos to spread malware to developers.

Impact: About 513,000 lines of Claude Code source exposed; developers who ran the fake leak archive infected with Vidar infostealer and GhostSocks proxy malware.

Who triggered it

Actor: Anthropic packaging error; GitHub user idbzoomh spreading Vidar and GhostSocks

Attribution: Leak self-disclosed by Anthropic as human error; malware lures identified by Zscaler and Trend Micro, operator not named

Motive: Leak accidental; follow-on campaign financially motivated credential theft and proxy botnet building

How it happened

Attack vector: Source map in npm package 2.1.88 exposing TypeScript source; SEO-optimized fake leak repos with a Rust dropper in GitHub Releases

Root cause: Build pipeline shipped a 59.8 MB source map; developers downloading unverified archives from repositories claiming to host the leak

Timeline

  1. Anthropic publishes @anthropic-ai/claude-code 2.1.88 with full source map
  2. Fake leaked Claude Code repos appear on GitHub pushing Vidar and GhostSocks
  3. Zscaler, Trend Micro and media report the malware campaign

Response

Anthropic pulled the package and called it a packaging error; researchers warned against downloading leak repos; malicious repos reported to GitHub.

Current status: Contained

Sources (open in a new tab)

Cite this case

Claude Code source map leak followed by malware-laced fake 'leak' repos. AI Threat Watch, case claude-code-leak-lures, Mar 31, 2026. https://aithreat.io/cases/claude-code-leak-lures/

Permanent link: https://aithreat.io/cases/claude-code-leak-lures/

This case file in other languages: EnglishFrançaisEspañolالعربية中文