A tampered Claude Code extension was available on the Cursor marketplace for about 11 hours. According to The Decoder, it could affect up to 5,000 users. The malicious code tried to steal user data and set up remote access to the machine where the extension ran.
The problem did not lie in the AI model itself but in the distribution channel. The extension looked like a legitimate tool supporting developer work, and it was installed from an official store, which normally builds trust.
Cursor took the entire marketplace offline and reinstated it only after confirming there was no further risk. This means the vendor treated the incident as a systemic issue, not as a single faulty package. For teams, it is a signal that even a trusted extension store does not guarantee that every listing is safe.
AI coding tools operate in developer environments where they often have access to source code, environment variables, API keys and repositories. An extension with such access can steal data or open a path to remote control of a workstation. The risk grows when developers install add-ons on their own, without approval rules.
Companies should introduce a simple process:
The Cursor case shows that extension marketplaces are part of the software supply chain. Companies using AI coding tools should treat extensions as potential entry points for data theft and remote access, and apply the same controls they use for other third-party code. The incident details come from The Decoder, so teams should also check them against the vendor's own communication before drawing final conclusions.
Lub System helps B2B companies implement AI, automation and IT solutions end-to-end - from strategy to deployment. See our services or get in touch to discuss your case.