AI adoption is accelerating across enterprises, with employees increasingly relying on platforms like ChatGPT, Claude, and Microsoft Copilot for everyday work.
Cybercriminals are exploiting this growing trust, distributing malicious applications disguised as legitimate AI tools to steal credentials, deliver malware, and gain access to corporate environments.
Recent research by ANY.RUN reveals how attackers are impersonating popular AI brands to target corporate users, including developers and marketing teams.
The findings highlight a growing challenge for security leaders: identifying threats hidden behind familiar applications before they lead to credential theft, compromised accounts, or wider business disruption.
Who Is Most Exposed to Fake AI Attacks?
Between April 15 and October 5, 2026, ANY.RUN researchers identified 1,258 public sandbox analyses linked to fake AI campaigns.
The United States accounted for 16% of submissions with country information, ranking second after India (20%). US submissions also led in two major campaign groups involving fake Claude downloads and ChatGPT-themed attacks.
The activity spans multiple industries. Technology appeared across all seven campaign queries examined by researchers, followed by telecommunications (six), education and consulting (five each), and manufacturing, healthcare, government, and finance (four each).
Among US submissions, telecommunications and manufacturing appeared most frequently across the campaign queries.
Attackers also tailor their methods to specific employee roles:
- Developers: Fake AI coding assistants and installation instructions can put source code, cloud credentials, and API keys at risk. In one sandbox session, researchers observed a counterfeit GitHub Copilot CLI page delivering malicious installation instructions.
- Marketing teams: Counterfeit AI advertising programs target employees with access to advertising budgets, billing systems, and business accounts. A fake Claude Ads campaign used a fraudulent Google sign-in window to capture credentials.
- Other corporate users: Fake desktop applications can expose employees to malware and browser-session theft. One trojanized Claude installer launched the legitimate application while secretly collecting browser data in the background.
For business leaders, these findings show that fake AI threats are not limited to technology companies or IT departments.
Exposure can arise wherever employees install AI applications or connect business accounts without adequate security oversight.
Reduce the Impact of Fake AI Attacks.
Detect malicious behavior earlier and contain threats faster. Strengthen Threat Response
How Fake AI Attacks Unfold: From Trusted Brand to Corporate Compromise
ANY.RUN’s research highlights how attackers move from AI-branded lures to malicious execution and potential corporate compromise:
- Initial lure: Attackers impersonate ChatGPT, Claude, Microsoft Copilot, and Google Gemini through counterfeit websites, advertisements, and shared AI conversations.
- Malicious delivery: Fake installers and setup instructions trick users into downloading malware or executing malicious commands disguised as legitimate installation steps.
- Payload execution: Once launched, malicious code can deploy information stealers or other malware on the employee’s device.
- Credential theft: Attackers attempt to collect passwords, session cookies, authentication tokens, and other sensitive data.
- Business impact: Stolen credentials and compromised sessions can enable unauthorized access to corporate systems, potentially leading to data exposure or further intrusion.
Fake AI Threats Extend Beyond Windows
Windows remains a major target for malware campaigns, but macOS devices are not immune.
Fake AI installation pages can trick Mac users into copying malicious commands into Terminal, presenting dangerous actions as routine software setup instructions.
For enterprises with mixed operating systems, this creates an additional investigation challenge. Limited visibility into macOS activity can leave security teams without the evidence needed to assess an incident, potentially delaying containment and leaving affected accounts or devices unidentified.
Cross-platform behavioral analysis helps address this gap. For example, ANY.RUN’s Interactive Sandbox supports Windows, Linux, Android, and macOS, allowing analysts to investigate suspicious files and URLs in isolated environments.
Teams can observe malicious behavior, examine network connections, extract indicators of compromise (IOCs), and trace attack chains to determine what happened and what requires further investigation.
With consistent investigation capabilities across operating systems, SOC teams can reduce blind spots, avoid fragmented investigations, and make more informed response decisions without relying on separate analysis workflows for every platform.
Closing the AI Security Visibility Gap
As AI adoption grows, enterprises need to ensure their security teams can investigate threats disguised as legitimate applications.
This means looking beyond suspicious downloads to understand what executes on employee devices, which credentials may be exposed, and whether an attack has progressed further into the corporate environment.
More than 16,000 organizations worldwide use ANY.RUN to investigate cyber threats and strengthen SOC operations.
Teams using the solution report up to 36% higher detection rates, 21 minutes less MTTR per incident, and 30% fewer Tier 1-to-Tier 2 escalations.
For enterprises, these improvements mean faster investigations, fewer unnecessary escalations, and more time for analysts to focus on critical threats.
As attackers continue to exploit trusted AI brands, the priority for security leaders is clear: give SOC teams the cross-platform visibility and behavioral evidence they need to identify malicious activity early and contain threats before they cause wider business disruption.
Help Your SOC Stop More Threats with the Same Team.
Cut investigation time and reduce escalations. Scale SOC Performance
