Fake AI Campaigns Target Businesses Through Malware and Credential Theft

Executive Summary

Cybersecurity researchers have identified a growing threat involving fraudulent websites and campaigns that impersonate popular artificial intelligence (AI) tools, including ChatGPT, Claude, Google Gemini and GitHub Copilot.

A recent research highlights how attackers exploit the popularity of AI applications to trick employees into downloading malicious software, running harmful commands or entering their login details on fake websites.

Research by ANY.RUN identified 1,258 public sandbox analyses linked to fake AI campaigns between April and October 2026. The United States accounted for 16% of targets attributed to specific countries in the reported dataset.

These attacks can lead to malware infections, stolen browser sessions, compromised accounts and unauthorised access to corporate systems. For organisations, the risk extends beyond a single infected device, as attackers may use stolen credentials to access email accounts, cloud platforms and sensitive business information.

How the Attacks Work

These campaigns rely on impersonation, social engineering and the misuse of legitimate system tools to make malicious activity appear trustworthy.

A reported campaign impersonating Claude demonstrates how the attack can unfold:

• Fake website: A user visits a fraudulent website designed to resemble the official AI application's download page.

• Malicious instructions: Clicking the download button copies a command to the user's clipboard and instructs them to run it through Windows PowerShell.

• Malware delivery: The command uses legitimate Windows utilities, including mshta.exe, to retrieve and execute malicious content.

• Malware execution: Encoded PowerShell commands help launch the malicious payload.

• Information theft: Once the device is compromised, the malware may steal login credentials, browser session data and other sensitive information.

Because users may be persuaded to run the commands themselves, traditional download filtering alone may not stop the attack.

Potential Business Impact

A successful attack may result in:

• Theft of employee credentials and browser authentication sessions.

• Unauthorised access to corporate email, cloud applications and business systems.

• Malware infections and persistent access to employee devices.

• Exposure of confidential company or customer information.

• Further compromise through stolen accounts and credentials.

Recommended Security Measures

Employee Awareness and Safe Software Downloads

Employees should be trained to recognise fake AI websites, misleading advertisements and fraudulent application downloads.

Endpoint Protection and Threat Detection

Endpoint detection and response (EDR) solutions should be deployed, updated and actively monitored across supported devices.

Identity and Data Protection

Organisations should enforce multi-factor authentication (MFA), prioritising phishing-resistant methods for privileged and sensitive accounts where available.

Security Monitoring and Threat Intelligence

Security teams should correlate related alerts across endpoints and environments to identify recurring attack patterns and reduce repetitive investigations.

Incident Response: What to Do if an Attack Is Suspected

If an employee runs a suspicious command or installs a suspected fake AI application, the security team should:

• Contain the threat: Isolate the affected device from the network where appropriate while preserving evidence for investigation.

• Investigate the activity: Review process execution, PowerShell logs, endpoint alerts, downloaded files and outbound network connections.

• Assess possible data exposure: Determine whether credentials, browser sessions or sensitive information may have been stolen.

• Secure affected accounts: Revoke active sessions, reset compromised credentials and investigate suspicious account activity using a trusted device.

• Check for further compromise: Search other endpoints and relevant environments for the same indicators and attack behaviours.

• Remove the threat and recover: If compromise is confirmed, remove the malware and restore the affected system using trusted recovery procedures.

Fake AI campaigns show how attackers can exploit familiar technology brands to trick users into running malicious commands and exposing sensitive information. Preventing these attacks requires more than blocking suspicious downloads; organisations also need employee awareness, endpoint monitoring, identity protection and effective incident response.

Remind employees to download AI tools only from verified sources, review endpoint alerts for suspicious PowerShell and mshta.exe activity, and investigate any suspected credential or browser-session compromise promptly.