34 Confirmed Cases of AI Brand Impersonation
Sophos X-Ops has confirmed 34 instances of AI brand impersonation used to deliver malware, spanning July 2025 through June 2026. In each case, attackers impersonated Claude, ChatGPT, or Copilot to trick users into executing a payload.
The delivery mechanism is a combination of malvertising and infostealers, which Sophos says is the primary initial access vector for The Gentlemen’s current campaign. The AI brand front lowers the friction threshold compared to traditional ransomware download links, and the infostealer logs feed directly into the credential pool that The Gentlemen uses for valid-account logins.
How the AI Brand Fakes Work
Victims are lured to a fake AI brand page via malvertising or an infostealer. The page mimics the branding of Claude, ChatGPT, or Copilot, depending on the target. Once the user lands, the malware is delivered and executed to complete the infection chain.
Who Was Hit and What They Got
Across the 12-month window, The Gentlemen confirmed 34 victims. The malware stack is a nine-tool set: Amadey, Backdoor.Turn, Edgecution, Fox Tempest, Qilin Encryptor, RTCore64.sys, SocGholish, StealBit, and StealC. The group references nine CVEs in its exploitation chain, spanning CVE-2023-20269 through CVE-2026-50751, which cover the AnyConnect SSL VPN, Bleed (NetScaler), Connect Secure, ESXi, FortiGate, GlobalProtect, RMM, Remote Access VPN, and driver components used for initial access and lateral movement.
Key points
- 34 confirmed cases across Jul 2025–Jun 2026
- Attackers fake Claude, ChatGPT, and Copilot UIs
- Delivery via malvertising and infostealer payloads
- No CVE referenced; pure brand spoofing + payload chain