Home / Technology / Claude AI Feature Abused for Infostealers
Claude AI Feature Abused for Infostealers
19 Jun
Summary
- Criminals exploit Claude AI's Shared Chats for malware distribution.
- Fake Apple Support chats promoted via Google Ads targeted developers.
- Malicious commands led to infostealer infections on macOS.
- Anthropic has since banned malicious accounts and disabled conversations.
Security researchers have identified a malicious campaign leveraging Claude AI's 'Shared Chats' feature to spread infostealer malware. Criminals created fake Apple Support conversations hosted on the claude.ai domain, shared via clickable links. These chats instructed users, primarily software developers on macOS, to execute commands in their Terminal.
These commands, part of a ClickFix scam, triggered a chain reaction leading to the installation of infostealer malware. The malicious chats were advertised through Google Ads, targeting users searching for terms like 'Claude Code on Mac.' While Russian-speaking regions were excluded, the majority of affected users were in Taiwan, followed by Japan, Singapore, and the US.
Anthropic, the AI company behind Claude, has investigated the abuse. The company has since banned the accounts responsible for the malicious activity and disabled the offending shared conversations. Anthropic has stated that additional measures are being implemented to prevent future exploitation of its platform.
