A real malware campaign called Yoba used fake GitHub projects advertising supposed ChatGPT and Claude desktop applications. Victims were lured by high-ranking Google results and download buttons, then ran ZIP-packaged installers that appeared legitimate but gave attackers control of the device. The campaign used trusted signed software to help hide the malware, making unofficial downloads especially risky.
The Yoba campaign relied on counterfeit GitHub projects that posed as desktop versions of popular AI tools, using names like GPT-5.6 Sol, GPT-6 Astra, and Claude Opus. These projects used GitHub Pages sites to describe the fake software and displayed download buttons that pointed to ZIP files stored in GitHub Releases. Inside the ZIP was a legitimate signed Setup.exe bundled with a malicious DLL. Once run, the malware established persistence through a scheduled task disguised as a Microsoft component and ultimately delivered an XMRig cryptocurrency miner, communicating through Tor.
The campaign exploited trust in two places at once: search engine rankings and GitHub itself. SEO poisoning tuned titles, descriptions, and hosting domains so the fake projects appeared near the top of Google results, even though they had no connection to OpenAI or Anthropic. Because many developers and knowledge workers already trust GitHub as a platform, the projects looked credible to visitors who assumed that ranking highly and being hosted on a well-known code repository implied legitimacy. Bundling a legitimate signed installer alongside the malicious DLL added another layer of apparent authenticity.
Organizations and individuals can reduce exposure to campaigns like Yoba by reinforcing a few habits. Download AI tools only from official vendor websites such as openai.com or anthropic.com, rather than from search results or third-party GitHub projects. Treat a high Google ranking as a discovery aid, not proof of safety, since SEO poisoning can place malicious pages above legitimate ones. Before trusting any GitHub-hosted tool, check the repository owner, account age, commit history, and whether it has genuine contributors. Finally, avoid running unfamiliar software with administrator rights and involve IT review when a tool requests elevated permissions, since limiting administrator access and blocking unknown software sources can reduce the damage from campaigns that reuse this approach under new branding as AI tools evolve.
Mirage safely runs attacks like this one against your own team, so you find out what happens before a real adversary does.
Attackers created fake GitHub projects advertising desktop apps named GPT-5.6 Sol, GPT-6 Astra, and Claude Opus, then used SEO poisoning so the pages ranked near the top of Google search results.
The ZIP file from GitHub Releases contained a legitimate signed Setup.exe alongside a malicious DLL, which established persistence disguised as a Microsoft component and ultimately delivered an XMRig cryptocurrency miner through Tor.
Download AI tools only from official vendor websites such as openai.com or anthropic.com, and before trusting a GitHub project, check the repository owner, account age, commit history, and real contributor activity.
Because many developers trust GitHub as a platform, the fake projects looked legitimate, and a high Google ranking was mistaken for proof of safety even though it resulted from SEO poisoning.
Google puts “GPT-6 Astra” near the top. Looks promising? That’s SEO poisoning. Yoba’s fake GitHub projects advertise GPT-5.6 Sol, GPT-6 Astra, and Claude Opus. Their download buttons lead to GitHub Release ZIPs. Here’s the trap: the ZIP pairs a signed Setup.exe with a malicious DLL. Run it, and Yoba can persist as a fake Microsoft task, then fetch an XMRig miner through Tor. A top result isn’t a trusted download. For AI apps, go directly to openai.com or anthropic.com.
Check Point Research identified a real campaign, tracked as Yoba, that uses fake GitHub projects and websites posing as new ChatGPT and Claude desktop applications. Victims searching for AI tools can be directed to convincing pages for supposed releases such as GPT-6 Astra and Claude Opus, then…
Sophos reports real incidents where attackers impersonated well-known AI brands (especially Claude) to trick people into installing malware. The lures included polished fake installer pages that instruct victims to copy/paste commands, and browser extensions that look legitimate via high ratings…
Three real malware campaigns targeted people in Brazil and Portugal with fake legal notices, invoice pages, and finance-themed documents. Victims were directed to scripts, installers, or an ISO file that abused trusted VLC, VMware, and Microsoft utilities to load malicious code. The most useful…
Researchers say a law firm was targeted with a spear‑phishing email that led staff to download an encrypted archive containing a Windows shortcut labeled like legal case files. After the user ran it and approved admin rights, the malware told Microsoft Defender to ignore a folder and a fake…
Researchers reported a real spear‑phishing intrusion against a law firm where attackers sent a message with a link to an encrypted archive. The archive contained a Windows shortcut (LNK) disguised as “Case Documents,” and running it launched a multi‑stage loader (“HollowFrame”) that ultimately…
This bulletin describes multiple real-world social engineering campaigns where attackers impersonate IT support or use trusted-looking sharing and “Allow” prompts to gain access. Several campaigns abuse Microsoft Teams and document-sharing lures to trick employees into installing remote tools or…
