Anamika Dey, editor
By TechSun News Desk | techsunnews.com | August 17, 2026 | Tech / Trending / Security | 6 min read
The tool you trust is becoming the disguise the attacker wears. Security researchers at Kaspersky have detailed a campaign in which a cyberespionage group distributes fake versions of Claude — Anthropic’s popular AI assistant — to slip malware onto company systems. And while this particular operation is concentrated in Asia, the tactic behind it is now a global problem worth every reader’s attention: fake AI apps have quietly become one of the most effective malware lures on the internet.
Here’s what Kaspersky found, why criminals are impersonating AI tools specifically, and — the part that matters most for you — how to make sure the AI app you download is the real one.
What Kaspersky found
According to Kaspersky’s Global Research and Analysis Team (GReAT), a threat group it tracks as SilverFox has been distributing counterfeit Claude applications built for Windows, macOS and Linux. The fake apps imitate Anthropic’s real chatbot, but once installed they quietly embed malware that gives the attackers long-term access to the machine and lets them harvest sensitive data. Kaspersky’s researchers presented the findings at the firm’s Cyber Security Weekend for the Asia-Pacific region.
Kaspersky says it first identified SilverFox in December 2025 and describes it as one of the most active threat groups in the Asia-Pacific region — responsible, by the firm’s estimate, for roughly 60% of the group’s global activity concentrated in East and South Asia. Named recent targets, according to the researchers, include companies in India, Indonesia, South Africa and Russia, across the industrial, consulting, trade and transport sectors.
Who is SilverFox?
Kaspersky classifies SilverFox as an advanced persistent threat (APT) group — the term for a well-resourced, patient operation focused on long-term espionage rather than a quick smash-and-grab. The researchers say the group customizes known malware families such as ValleyRAT and Winos to slip past standard antivirus, and spreads them through phishing emails, compromised websites and malicious server addresses. In one strand of the campaign, Kaspersky logged more than 1,600 malicious emails between January and February 2026, many disguised as official tax-audit notifications urging the recipient to open an attached “list of tax violations.”
As one Kaspersky researcher put it at the event, the attacks almost always come down to a few simple methods — ordinary phishing, phishing through trusted websites, and attacks via malicious addresses — dressed up in whatever lure is most likely to get a click. Right now, that lure is increasingly AI.
Why criminals are impersonating AI apps
The logic is grimly simple. Millions of people are downloading AI tools for the first time, often without knowing the correct official website — so they search, and they trust whatever looks right. That’s a perfect opening for an attacker. And the scale is no longer small.
| 📊 This is Bigger than one group
The SilverFox campaign is one example of a much larger trend. Kaspersky says that between January and mid-2026 it detected more than 92,000 malware attacks worldwide disguised as popular AI services. Fake ChatGPT apps accounted for roughly 49% of them; fake Claude and fake Gemini apps made up about 18% each. The firm also found over 15,000 malware samples impersonating newer “agentic” AI tools like OpenClaw. In other words, impersonating an AI app is no longer a niche trick — it has become one of the internet’s most common malware delivery methods. |
This is exactly the risk we flagged in our guide to AI scams to watch for in 2026 — fake AI apps and downloads were one of the core categories then, and this campaign is that warning playing out at industrial scale.
How to Download an AI app safely (the reader takeaway)
You don’t need to avoid AI tools. You need to download them the right way. A few habits defeat almost every version of this scam:
- Go to the official website directly. Type the real address yourself or use the official app store listing. For Claude that’s claude.ai / Anthropic’s official site; for others, the vendor’s own domain.
- Never download an AI app from a search ad. Attackers buy sponsored results that sit above the real one. Scroll past the ads to the genuine site, or better, type the URL yourself.
- Be suspicious of emailed “downloads.” A legitimate AI tool doesn’t arrive as an attachment or a link in an unsolicited email — especially one creating urgency, like a tax or compliance notice.
- Check the details. Misspelled domains, a download hosted on a random site, missing developer information, or an installer that asks for unusual permissions are all red flags.
- Keep security software on. It won’t catch everything, but it catches a lot — and pair it with knowing the signs that a device has already been compromised.
What businesses should take from this
For companies, the uncomfortable lesson is that the danger now rides in on a productivity tool employees genuinely want to use. Kaspersky’s own guidance points the same way: give staff an approved, official source for AI tools so they’re not hunting through search results; train them to recognize AI-app phishing lures; and treat AI software downloads with the same scrutiny as any other install. The alternative is an employee innocently downloading “Claude” and handing an APT group a foothold.
The bottom line
The specific victims here are mostly in Asia, but the technique is borderless and the takeaway is universal: as AI tools become something everyone downloads, impersonating them has become something attackers do at scale. The fix isn’t fear of AI — it’s the same discipline that protects you everywhere online. Get your software from the source, distrust the shortcut, and slow down when something urgent tells you to click. It’s the same lesson our recent piece on an AI agent exploiting a weak system pointed to from a different angle: the technology keeps changing, but the weak link is almost always the same old human click.
| 💬 Over to you
How careful are you about where you download AI apps?
Tell us in the comments — and share this with anyone who installs AI tools without thinking twice. |
Frequently Asked Questions
Are fake Claude apps a real threat to me?
The specific campaign Kaspersky detailed targets businesses, largely in the Asia-Pacific region, so most individual users are not the direct target. But the underlying tactic — malware disguised as a popular AI app — is worldwide, and Kaspersky says fake ChatGPT, Claude and Gemini apps together accounted for tens of thousands of attacks in 2026. Anyone who downloads AI tools should take the same precautions: get them only from the official source.
How do I know if a Claude or AI app is the real one?
Download only from the developer’s official website (typed directly, not via a search ad) or an official app store listing. Check for a correctly spelled domain, verified developer information, and reviews. Be especially wary of AI apps arriving as email attachments or links, installers hosted on random websites, or any download that pressures you to act urgently — all common signs of the fake-app scams researchers have documented.
What is SilverFox?
SilverFox is the name Kaspersky uses for an advanced persistent threat (APT) group it says it first identified in December 2025 and considers one of the most active in the Asia-Pacific region. According to Kaspersky, the group focuses on long-term cyberespionage and data theft, customizes malware such as ValleyRAT and Winos to evade antivirus, and has used fake Claude applications and tax-themed phishing emails as part of its attacks.
| ✍️ Editor’s Observation
There’s a bitter irony in criminals impersonating Claude and ChatGPT specifically. These tools earned their popularity partly by feeling trustworthy — friendly, helpful, safe. That trust is now the attack surface. The reason a fake AI app works isn’t clever code; it’s that we’ve been trained to want these tools and to grab them quickly. That’s worth sitting with for a second the next time you go to install one. The safest move is almost boringly simple: decide where the real download lives before you search, go straight there, and ignore everything that offers you a faster route. In security, the shortcut is almost always the trap. |
— TechSun News, Security Desk
