Fake ChatGPT and Claude apps install miners: key points for your SME

The artificial intelligence ecosystem has become a perfect lure for cybercrime. According to The Hacker News, a campaign dubbed Yoba is distributing malicious loaders and cryptocurrency miners disguised as desktop applications from Anthropic Claude and OpenAI ChatGPT. The modus operandi: creating fake projects on GitHub that present themselves as supposed desktop apps with names like GPT-5.6 Sol, GPT-6 Astra or Claude Opus, and that actually deliver a malware family whose goal is to mine cryptocurrency on the victim's machine.

The detail that raises the severity of the matter is the use of signed executables. The attackers abuse binaries with a legitimate digital signature to bypass security checks and give the user an appearance of trust. In other words, this is not a simple suspicious installer that antivirus detects right away, but rather an attack chain designed to look legitimate in an environment where AI generates almost automatic trust.

Why this campaign should worry an SME

The temptation to download "the ChatGPT desktop app" or "the Claude client" is very high in small teams looking for quick productivity. And that is the problem: many users do not distinguish between the manufacturer's official website and a GitHub repository with an attractive name. The Yoba campaign exploits exactly that ambiguity.

For an SME, the impact is not only the miner. A machine compromised by a cryptocurrency miner usually entails high electricity consumption, performance degradation, hardware overheating and, above all, an open door to other threats. Miners rarely arrive alone: the loader that installs them can later download other payloads, steal credentials or serve as an entry point to the corporate network. In a company with few IT resources, detecting that anomalous consumption without monitoring is almost impossible until the damage is done.

The pattern repeats: GitHub as a shop window

This is not the first time the code repository has become a distribution vector. We recently analyzed how GitHub is shielding itself for the era of agents, and the Yoba campaign shows why that protection is necessary. Attackers know that a project with stars, a careful README and a name that sounds like an official product can fool anyone. The difference now is that the bait is AI, the topic that generates the most interest among companies.

In addition, the use of signed executables connects with another trend we have already seen: the trust placed in the digital signature as an absolute guarantee. It is not. A valid signature only indicates that a binary has not been modified since it was signed, not that its purpose is legitimate. If the attacker manages to sign their own executable or abuse a signed one, the security barrier blurs.

What we recommend from ForgeNEX

Faced with this type of campaign, the answer is not to stop using AI, but to adopt specific habits and controls. These are the measures we advise our SME clients:

• Download only from official channels. OpenAI and Anthropic desktop applications should be obtained exclusively from their official websites. If a GitHub project is not linked from the manufacturer's website, it is not official.
• Be suspicious of names that sound like products. GPT-5.6 Sol, GPT-6 Astra or Claude Opus may look like new versions, but if they do not appear on the manufacturer's website, they are bait. Always verify the source before running anything.
• Control what is installed on machines. Applying an installation policy that requires IT approval for uncatalogued software drastically reduces the attack surface.
• Monitor consumption and performance. A miner gives itself away through sustained CPU and GPU use. Basic monitoring tools make it possible to detect anomalous spikes outside working hours.
• Do not blindly trust the digital signature. Strengthen binary analysis with EDR solutions that evaluate behavior, not just reputation.
• Train your team. Awareness remains the cheapest defense: explaining that a fake AI app can be costly prevents many incidents.

The Yoba campaign is a reminder that the popularity of AI also attracts those seeking quick profit. For an SME, the key is to treat any AI-related download with the same rigor as financial software: verify origin, control installation and monitor subsequent behavior. If you need help reviewing your download channels or strengthening endpoint monitoring, at ForgeNEX we can help you set up that control without blowing the budget.

Source: The Hacker News. Analysis and adaptation: ForgeNEX.