Fake AI Download Pages Are Tricking Employees Into Installing Malware

Attackers are posing as popular AI products to distribute malware, steal browser sessions and compromise business accounts. Security firm ANY.RUN linked 1,258 public sandbox analyses to fake AI campaigns between April and October of 2026. US businesses made up 16% of country-attributed targets. The lures impersonate Claude, ChatGPT, GitHub Copilot, Gemini and other services, and appear in sponsored search results, legitimate hosting services, shared AI conversations and counterfeit download websites.

One campaign analysed by ANY.RUN imitated the official Claude download site. When the victim clicked the Windows download button, a command was copied to their clipboard and the page told them to run it in PowerShell. Instead of installing the app, the command launched a built-in Windows tool, retrieved a malicious file and started a chain of hidden PowerShell steps. This is hard to detect because the employee starts the process themselves, the attack uses legitimate Windows utilities, and the harmful payload only appears after several stages.

The source article focuses on the strain these campaigns place on managed security service providers (MSSPs). Similar incidents across different customers can trigger repetitive investigations, use up analyst time and slow response. Other lures push employees to install compromised applications or enter credentials on fraudulent login pages.