Skip to the content
AI and Cybersecurity Learning and Practice Center Test automation Open the Learning Center

NewsAnalysis

Fake Claude and ChatGPT installers: 1,258 incidents in six months, with Claude impersonated in 60 percent and a pasted command in nearly half

ANY.RUN traced 1,258 incidents on its sandbox platform over six months to fake AI tool lures, Dark Reading reported on October 16. Claude was impersonated in 60 percent of the analyses and ChatGPT or Codex in 33 percent. Tactics of the ClickFix family made up nearly half: 45 percent of analyses had the victim copy and paste a command, fake command-line installers that ANY.RUN calls InstallFix were 27 percent, fake download pages 17 percent and abused AI chats or malicious artifacts 25 percent. The lures ride SEO poisoning and malicious ads, and are hosted as fake installers on GitHub or as copied documentation on Cloudflare Pages, GitLab Pages, Google Sites and the AI platforms themselves; Push Security's Dan Green describes LLMshare, malicious shared chats and artifacts on legitimate AI domains that offer installation help, at 5 to 10 percent of its monthly detections and growing. Developers lose source code, SSH keys, CI/CD tokens and cloud and AI API keys; marketing staff lose ad and billing accounts. ANY.RUN could not say how many infections resulted.

Why it matters

The install command a colleague pastes from a search result is the attack. Publish the list of approved AI tools with their official sources, install them through your software portal, and teach people to check an install command against the vendor's own documentation.

Sources

Read it with the rest of the news in the Learning Center