How Hackers Use Fake Ai Tools to Spread Malware Through Social Media Ads

Cybercriminals are hijacking the global fascination with artificial intelligence to run a massive online scam, using fake social media ads to spread malware and steal personal data from unsuspecting users.

A new report by cybersecurity firm Mandiant has exposed a campaign by a Vietnam-linked group known as UNC6032, which has been operating since at least mid-2024. The group is using fake websites that imitate well-known Ai tools — including Luma Ai, Canva Dream Lab and Kling Ai — to trick people into downloading malicious software.

What’s happening?

The ads, which have appeared on Facebook and LinkedIn, promote what look like AI-powered video or image generators. But when users click through, they’re taken to a spoofed website that mimics the appearance of a legitimate AI tool.

Instead of producing any AI-generated content, the website silently installs malware that can steal login credentials, payment information and other sensitive data. The stolen data is then exfiltrated via Telegram and likely sold on underground markets, the report says.

How big is the threat?

Mandiant has identified thousands of malicious ads, with a sample of 120 Facebook ads alone reaching an estimated 2.3 million users across the European Union. Many of the ads were pushed out via compromised accounts or fake business pages, and new ones are being launched daily to avoid detection.

LinkedIn was also targeted, with a handful of ads promoting a fake site called klingxai[.]com reaching tens of thousands of users.

This isn’t just a phishing scam — it’s a professionally run malware campaign exploiting the AI boom. It’s a risk to individuals and organisations alike.

The malware being deployed — known as STARKVEIL — is modular and designed to steal information, maintain persistence, and avoid detection. Mandiant’s researchers say the use of multiple overlapping payloads is a fail-safe measure to ensure the attack succeeds even if some parts are blocked.

How are tech companies responding?

Mandiant collaborated with Meta and LinkedIn, who have taken down many of the ads, accounts and domains involved. Meta’s threat detection efforts were already underway before Mandiant flagged the latest wave of activity.

But the report warns that new domains and ads are being created daily, highlighting the need for ongoing industry coordination and public vigilance.

What can you do?

Experts recommend:

• Avoid clicking on AI-related ads unless you’re sure of the source
  
  
• Stick to verified websites for downloading tools or apps
  
  
• Use antivirus software and enable browser protection features
  
  
• Report suspicious ads or pages directly to the platform

The Mandiant report also notes that stolen credentials are now the second most common method of cyber intrusion, making awareness and prevention more important than ever.

Ai isn’t the threat, it’s the way attackers are using the excitement around AI to exploit trust — and that’s what we all need to watch out for.

Yash Gupta is Senior Manager at Mandiant Threat Defense.

Last Updated on May 30, 2025 by Yash Gupta