Fake Claude and ChatGPT lures appear in 1,258 public sandbox analyses from April to October
ANY.RUN tied 1,258 public sandbox analyses to fake AI-brand campaigns between April and October 2026. The vendor says repeat cases strain MSSP analyst capacity, though the article does not report analyst hours.
The Watch · Security desk
Key figures: US businesses accounted for 16% of country-attributed targets, per ANY.RUN; 84% of country-attributed targets were outside the US; ANY.RUN linked 1,258 public sandbox analyses to fake AI campaigns between April and October 2026.
| Figure | Value | Note | Claim |
|---|---|---|---|
| US businesses' share of country-attributed targets | 16% | 3 | |
| Targets outside the US | 84% attributed | 9 | |
| Public sandbox analyses linked to fake AI campaigns | 1,258 analyses | April to October 2026, per ANY.RUN | 1 |
What happened
- The lures impersonate Claude, ChatGPT, GitHub Copilot, Gemini and other AI services, and appear in sponsored search results, shared AI conversations, legitimate hosting and counterfeit download sites.
- In the Claude case ANY.RUN analyzed, the fake download page copies a command to the clipboard when the user clicks the Windows button, then shows instructions to run it in PowerShell.
- The pasted command launches mshta.exe, retrieves a malicious HTA file and starts an encoded PowerShell execution chain instead of installing any application.
Why it matters
- What it limits constraint The employee starts the chain voluntarily, it runs on legitimate Windows utilities, and the payload appears only after several stages, so the first alert often shows little that looks malicious.
- What it costs cost MSSP analysts bear the repeat work if the same chain shows up at several customers, because ANY.RUN says each case can trigger its own investigation and lengthen response times.
- Who has to decide decision An MSSP weighing ANY.RUN's sandbox is looking at the vendor's own figures of 30% fewer Tier 1 to Tier 2 escalations and a 14-second median detection time, which it should test on its own queue.
The mshta chain is one route. ANY.RUN also describes lures that send employees to fraudulent authentication pages to submit credentials [8], and it lists the goals as malware delivery, browser-session theft and business-account compromise [7].
For an MSSP the useful detail is what stays fixed. According to the article, the initial indicators vary from case to case while the underlying attacker behavior remains similar [12]. Attackers rotate domains, hosting services and AI brands, which makes new variants harder to catch with existing indicators [13]. In our view that argues for detections on the process chain, mshta.exe retrieving an HTA file and handing off to encoded PowerShell, because the domains and brand names change on the attacker's schedule.
The scale evidence is thin. The 1,258 figure counts public sandbox analyses [1]. The article does not break it down by campaign or brand, and it does not say how many of those analyses involved MSSP customers.
The geography is also narrow in what it supports. US businesses were 16% of country-attributed targets [3], which leaves 84% outside the US [9]. An MSSP with US-only clients sees a minority of the targeting ANY.RUN describes.
Everything here comes from one publisher, and the piece closes on a product pitch for ANY.RUN's own sandbox. The campaign description is detailed enough to act on. The claim about analyst workload is the vendor's assertion [11].
What to watch
- Whether ANY.RUN publishes a breakdown of the 1,258 analyses by campaign, brand and delivery channel.
- Whether any MSSP reports measured analyst hours per fake-AI case, which would test the capacity claim.
- Whether ad platforms remove the sponsored search results that carry the counterfeit Claude and ChatGPT download pages.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence40
- Adoption
- Insufficient
- Hype gap+30
- Incentives80
- Confidence55
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
ANY.RUN linked 1,258 public sandbox analyses to fake AI campaigns between April and October of 2026.
- [2]
Attackers impersonate Claude, ChatGPT, GitHub Copilot, Gemini and other AI services, with lures in sponsored search results, legitimate hosting services, shared AI conversations and counterfeit download websites.
- [3]
US businesses accounted for 16% of country-attributed targets.
- [4]
In the Claude campaign analyzed in ANY.RUN's Interactive Sandbox, the victim lands on a page resembling the official Claude download site; clicking the Windows download button copies a command to the clipboard and displays instructions for executing it through PowerShell.
- [5]
Instead of installing the expected application, the command launches mshta.exe, retrieves a malicious HTA file, and triggers an encoded PowerShell execution chain.
- [6]
The employee initiates the process voluntarily, the attack relies on legitimate Windows utilities, and the malicious payload appears only after several execution stages.
- [7]
Attackers impersonate popular AI products to distribute malware, steal browser sessions and compromise business accounts.
- [8]
The lures encourage employees to execute malicious commands, install compromised applications, or submit credentials through fraudulent authentication pages.
- [9]
84% of country-attributed targets were outside the US.
- [10]
ANY.RUN reports 30% fewer Tier 1 to Tier 2 escalations when using its sandbox, alongside a median threat detection time of 14 seconds.
- [11]
Similar attacks appearing across different customers can trigger repetitive investigations, consume analyst capacity and extend response times for MSSPs.
- [12]
For MSSPs managing multiple customer environments, the initial indicators vary while the underlying attacker behavior remains similar.
- [13]
Attackers constantly rotate domains, hosting services and AI brands, making new campaign variants harder to detect with existing indicators.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- thehackernews.comFake AI Attacks Spread Across US Companies: How MSSPs Can Stop Incidents Early
1 article · October 15, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Malware sandbox analysisFollow
- AI brand impersonation malwareFollow
- Managed security servicesFollow