Christian Perry is the CEO of Undetectable AI and TruthScan, a deepfake-detection platform.
Thanks to deepfakes, companies have reportedly lost millions of dollars, including one instance of deepfake fraud that cost a company $25 million. As someone who works in the deepfake detection industry, I’ve seen firsthand how criminals are wreaking havoc on businesses and using deepfakes to do it.
Here are three ways criminals are commonly using AI deepfakes for fraud and what you can do to act fast before it’s too late.
1. Executive Impersonation
A familiar face or voice can make a fraudulent request seem legitimate. In 2024, engineering firm Arup lost approximately $25 million after an employee joined a video call with what appeared to be senior colleagues. The people on the call were deepfakes, and the employee transferred money to criminals.
Sometimes executive impersonation targets the public. An infamous example in recent years saw scammers using a deepfake video of Elon Musk to promote a fraudulent Neuralink-branded cryptocurrency scheme. These schemes exploit an executive’s reputation to make an unrelated offer appear to have the company’s backing.
How To Respond
Before acting on an unusual payment request, independently confirm it using contact information you already trust (even when you recognize the person’s face or voice). The FBI recommends verifying requests through a previously confirmed channel.
Businesses should also establish a response process for fake endorsements. If this happens to your org, preserve evidence, report the impersonating content to the platform and clarify through official channels that the offer is fraudulent.
2. Deepfake Documents
AI tools make it easier to create convincing fake documents, giving businesses another reason to diligently check the information behind an application.
What Deepfake Document Fraud Looks Like
Fraudulent job applicants can use AI to create résumés with invented work histories, fake credentials and false identity documents. For businesses handling large volumes of applications, those claims can often be difficult to verify at scale.
A document can look professional and still contain false information. Human review matters, but checking how a document looks is only part of the process.
How To Spot Deepfake Documents
Start by checking the claims. Verify degrees and certifications with the issuing institutions, and confirm employment history using contact information you find independently.
Document analysis software can help investigate suspicious files and look for visual AI patterns and indicators. You can also manually inspect available metadata, such as the author, creation date and last-modified date. These details may reveal inconsistencies that warrant a closer look.
Treat metadata as supporting evidence. It can be edited or removed, so it doesn’t establish AI authorship or fraud on its own. Whether someone used AI to prepare an application, you still need to verify that their identity and qualifications are real.
3. AI Refund, Return And Replacement Fraud
Whether your business offers goods or services, you could be vulnerable to AI-enabled refund fraud. Retail returns were estimated at nearly $850 billion in 2025, including roughly $76 billion in fraudulent returns.
What AI Refund Fraud Looks Like
We’ve seen scammers use AI-generated images of damaged products to justify fraudulent claims. They do this by making an item look hazardous or impossible to return, trying to get a full refund or replacement without having to send it back.
Businesses that sell services can also be targeted. Scammers can use AI to fabricate emails, text messages or agreements to support a fraudulent chargeback.
How To Prevent AI Refund Fraud
Document everything. Keep records of orders, deliveries and customer communications. If you sell physical products, make sure you or your team reviews the photos attached to refund claims and checks them against the order details.
Review the available image metadata, and use detection tools as an additional check if you suspect fraud. As an example, Google’s Gemini app can check for SynthID watermarks that identify images created or edited with Google AI. A missing watermark doesn’t rule out AI use, and a detection result should be considered alongside other evidence.
If a claim is still unclear, ask the customer for a video showing the damage and explaining what happened. Review it alongside your records and the customer’s photos. A longer video may give you more detail, but its length alone doesn’t prove it’s authentic.
Closing Thoughts
Deepfakes are making fraud harder to spot. Although the images and documents are fake, they can cause real businesses real harm. Before approving a payment, hiring an applicant or issuing a refund, take the extra step to verify what you’re seeing.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?