While AI arms fraudsters, transparent AI stops them

By Armen Najarian, CMO of Sift

Generative AI has become a defining force in today’s tech landscape, making everything from customer service to content creation faster and more dynamic. But the same technologies are also arming cybercriminals with scalable, adaptive tools to deceive, impersonate, and exploit. Fraud is no longer a matter of isolated attacks; it’s a complex web of automated, coordinated threats that change as quickly as the technology enabling them.

According to Sift’s Q1 2025 Digital Trust Index, payment fraud attack rates remain high at 3.3%, with sectors like fintech and online marketplaces seeing some of the highest activity. Meanwhile, the total cost of digital commerce fraud is expected to more than double by 2029, hitting $107 billion. Against this backdrop, businesses must rethink what effective fraud prevention really looks like and weed out the AI solutions that overpromise and undersell.

The new standard: Real-time identity trust

Stopping fraud today requires more than spotting fake credit cards or flagging suspicious IP addresses. It demands building trust and a deeper understanding of identity – the ability to evaluate whether a user’s behavior, context, and history indicate they are who they say they are. This means looking at signals over time, not just isolated events in order to strengthen decision confidence. This is the foundation for digital identity trust.

Identity trust becomes even more powerful when it extends across ecosystems. Today’s fraud isn’t siloed to one platform, it’s fluid and networked. If a bad actor has triggered fraud signals on another digital platform, cross-dimensional identity insights can expose them before they strike again. By surfacing patterns across a broader network, businesses can act earlier and more decisively, without adding unnecessary friction for legitimate users.

The challenge? Many companies still rely on fraud tools that function like black boxes: issuing decisions without offering any real explanation. These systems can be difficult to adjust, and even harder to trust. When a transaction is denied or approved, teams are left guessing why. This lack of transparency makes it nearly impossible to spot trends, respond to evolving threats, or fine-tune fraud policies without unintended fallout.

False declines are the hidden cost

Poor fraud decisioning doesn’t just let bad actors slip through; it blocks good customers too. When legitimate transactions are wrongly flagged as fraudulent, these false positives chip away at customer loyalty, lower conversion rates, and quietly drain revenue.

If fraud decisioning operates without context and ignores how trustworthy a user has proven to be over time – or how they’ve behaved across the broader digital ecosystem – businesses will continue to lose the confidence of consumers and shoppers. That’s why solutions that support dynamic, behavior-based decisioning and leveraging identity signals across a global data consortium are essential.

Why control matters more than ever

Static rules and rigid risk thresholds aren’t enough. Fraud patterns mutate constantly, especially with AI now being used to personalize scams and mimic legitimate users more effectively. The ability to adapt your fraud strategy in real time is no longer a bonus – it’s a requirement.

Look for systems that offer:

Full decision transparency: Understanding why a transaction was approved or denied – whether due to velocity, location, or device signals – helps teams fine-tune strategies, investigate manual review cases, and avoid unnecessary friction. Transparency fosters smarter decisions and greater trust in your system.

Dynamic fraud defense: Adapting to fluctuations in fraud patterns, seasonality, and geography requires systems that automatically adjust risk thresholds to maintain consistent outcomes, like block, friction, and review rates. Replacing static rules with intelligent automation allows teams to stay focused on strategy while ensuring effective protection against evolving threats.

Granular behavioral insights: Sophisticated fraud often hides in plain sight. Tracking identity behavior over time and across platforms, like logins, devices, or purchase patterns, helps detect threats such as account takeovers and payment fraud before damage is done.

Ensemble risk modeling: Combining risk models can be a key advantage for fraud teams. By applying three independent models (company-specific, industry-focused, and global consortium), companies can enhance fraud decision accuracy, reduce false positives and negatives, and protect against new threats without compromising legitimate user experiences.​

From cost center to growth enabler

Historically, fraud prevention has been viewed as a reactive safeguard, but that mindset is now outdated. When you can identify and trust legitimate users, you reduce unnecessary friction, increase merchant acceptance, and build long-term loyalty.

That’s the real shift: moving fraud prevention out of the shadows and into the strategic core of a growth-focused business. Instead of outsourcing control to black box systems, companies that embrace identity intelligence and demand visibility from their fraud stack will be better positioned to grow with confidence, no matter how the threat landscape evolves. This is the promise of fraud decisions driven through the lens of digital identity trust.

About the author

Armen Najarian is the CMO of Sift, the AI-powered fraud decisioning company.

Article Topics

AI fraud  |  biometrics  |  digital identity  |  digital trust  |  fraud prevention  |  generative AI  |  Sift