Fake installers for ChatGPT and Claude—properly signed and ranked at the top of Google search results—deliver malicious software: the Yoba campaign, discovered by Check Point.
Many employees search for AI apps directly on Google and install the first result. Attackers have capitalized on exactly this: they’ve published fake download pages for well-known AI assistants, hosted on GitHub, and pushed them to the top of search results through search engine optimization.
The surprising part: the installer contains a file that is properly signed by OpenAI or Anthropic, but alongside it, the attackers slip in a malicious component that loads along with the legitimate program and bypasses the signature verification. Once installed, the program creates a hidden scheduled task, disguised as a Microsoft task, and runs with full privileges.
For now, it uses the computer to mine cryptocurrency, but the download mechanism is generic and could deliver malware designed to steal or lock data at any time. The lesson for a small business: a digital signature alone is no longer proof that a program is clean, and the top result on Google could be a trap.
What you need to do:
- Download AI applications only from the manufacturer's official website (openai.com, claude.ai), never from search results, ads, or GitHub repositories.
- Determine who is authorized to install software on company computers; everyone else must request approval.
- If an AI application from an unknown source has already been installed, ask your IT provider to check for suspicious scheduled tasks running with system privileges and to perform a full antivirus scan.
No compromised password is needed: the attack relies on trusting the first result on Google and a valid signature.
Source: Check Point via Infosecurity Magazine | https://www.infosecurity-magazine.com/news/attackers-hide-malware-behind/
This article was generated with AI assistance.