24/7 local support: 480-550-8108Existing client? Just call, no ticket needed.
Cybersecurity

Fake ChatGPT and Claude desktop apps are spreading malware

Researchers found fake ChatGPT and Claude apps ranking high in Google search and hiding malware next to real, signed installers. How the trick works and how to keep your team safe.

By the Inventive Tech teamReviewed by our CISSP-certified security lead3 min readNovember 2026

More employees are using AI tools every month, and attackers have noticed. Researchers at Check Point recently reported a campaign they call Yoba, which hides malware inside fake desktop versions of ChatGPT and Claude. The fake downloads had been up since late August.

How the trick works

The attackers set up GitHub projects and download pages for apps with names like "GPT-6 Astra," "GPT-5.6 Sol" and "Claude Opus." The pages ranked high in Google search results, so someone looking for a ChatGPT or Claude desktop app could land on one without ever getting a suspicious email.

The download is a ZIP file with a real installer inside, digitally signed by OpenAI or Anthropic. That's what makes it convincing, because people and security tools both tend to trust signed software. Next to it sits a malicious file disguised as a standard Windows file, which the real program loads when it runs. There's no sign that OpenAI or Anthropic were hacked. The attackers reused their genuine, signed installers.

What it does

Once it's running with administrator rights, the malware sets up a scheduled task, disguised as a Windows Media Player task, so it starts again after every restart. It then fetches its next stage through Telegram, Google's DNS service and a file-sharing site, which helps it blend in with normal internet traffic.

In this campaign, the final payload was cryptocurrency mining software. It uses the computer's processor while nobody is at the keyboard and pauses when someone is, which makes it hard to notice. Check Point warns that Yoba "should not be viewed simply as cryptomining malware," because the same loader can download and run any other program, including ones that steal data.

How to keep your team safe

  • Download AI apps only from the source. Get ChatGPT from OpenAI's website and Claude from Anthropic's website, or from an official app store. Don't use links from search results, GitHub, forums or ads.
  • Make it a rule to ask before installing software, and back it up by limiting who has administrator rights. This malware needs administrator access to do its damage.
  • Keep a list of approved AI tools. Employees will use AI either way. Telling them which tools are approved, and where to get them, works better than a ban nobody follows.
  • If someone installed one of these apps, disconnect the computer from the network and call your IT provider. Treat it as a full compromise, and change any passwords used on that computer.

How we help

Our clients' computers run security software that watches for this kind of behavior, and we can limit administrator rights so unapproved programs can't install themselves. We also help businesses choose AI tools and set simple rules for using them. Our cybersecurity services and AI automation services cover the details.

Request a free IT assessment.

Here's exactly what happens when you request one:

  1. 1
    A 15-minute call about your business and how your IT works today.
  2. 2
    A review of your systems, security and backups, with your permission.
  3. 3
    A short, plain-English summary of your biggest risks, what an auditor or insurer would flag, and where you may be overspending.

Free, with no obligation. After you submit the form, we call you back the same business day to set up the first call.

Prefer to talk now? 480-550-8108

We only use your details to respond to your request. Privacy Policy

Call 480-550-8108Free assessment