Thanks to AI, fraud is rapidly rising and becoming more global in scope. Criminals are also shifting tactics by bypassing technological firewalls and targeting people directly, according to two recent research reports on the topic.
And the attacks are costly. In the U.S., Microblink estimates the cost of fraud attacks to reach $40 billion this year.
Microblink’s “Mapping the Rise of AI-Powered Identity Fraud” explored the evolution of deepfakes and their impact while Entrust’s “The Changing Face of Fraud: 2026 Identity Fraud Report,” examined the scope of the problem and how fraudsters are using AI to scale their activities.
“Deepfakes are no longer limited to manipulated videos,” said the authors of the Microblink report. “Advances in generative AI have made it possible to create highly realistic synthetic content across multiple modalities, including text, voice, images, documents and video. More importantly, these attacks can now be generated and deployed at an unprecedented scale. Just as phishing became profitable through automation and volume, AI is enabling fraudsters to industrialize identity fraud, social engineering, account creation and impersonation attacks.”
Deep fakes are the new norm, and their spread is being done automatically. Just like the rise of “agentic commerce,” there is now “agentic fraud,” and the implications for retailers and brands run deep.
For retailers and brands, the implications include direct attacks on their customers that can severely damage the relationship by souring the shopping experience. The aim is to mitigate fraud and cyber attacks by using various technologies and tactics while ensuring a smooth customer experience.
In the Entrust research report, the authors found that fraud tactics are shifting. “As detection systems improve at stopping sophisticated threats like deepfakes, attackers are finding success by targeting the people behind the technology,” they said. “Social engineering is also up year-over-year and presents a concern in the identity space. Due to the nature of this type of fraud, it’s tough to quantify. Still, coercion, phishing and impersonation scams are harder to stop because victims are convinced—or forced—to use their own genuine identity credentials.”
The Entrust researchers also found that fraudsters are using AI to scale their attacks. The research revealed that deepfakes are linked to one in every five cases of biometric fraud.
The Entrust report also found that attacks occur at all stages of the customer lifecycle and also change by industry. In crypto, for example, 67 percent of attacks occur at onboarding and are often driven by sign-up bonuses. In the payments space, 82 percent of fraud attempts target the authentication process, while digital banking sees 55 percent of fraud attacks occurring after onboarding.
To help understand the nature of the attacks, researchers at Microblink created three personas to categorize how fraudsters operate: the digital ghost, the master forger and the analog architect. Each work differently and varies according to geography.
“Depending on where your users are, the threat changes completely,” the report’s authors said. “What looks like a single global fraud problem is really a shifting cast of characters, each adapting to local conditions like a chameleon changing colors mid-stride.”
The research found that in some regions, attackers fully deploy digital deception by exploiting the gap between what a camera sees and what a human assumes is real. In other regions, the type of attack “moves closer to the document itself, with precision edits that turn legitimate IDs into near-perfect impostors.”
The report stated that in many parts of the world, the playbook isn’t futuristic at all. “It is grounded in physical reproduction techniques that exploit how often businesses still rely on visual inspection or low-fidelity capture.”
When it comes to the most established type of identity theft, document fraud reigns. By document type, national identity cards are number one, followed by a driver’s license and then a passport in third.
