Your fraud team is investing in artificial intelligence to catch criminals faster. But the criminals are investing in the exact same technology to stay a step ahead. AI and fraud have become inseparable—on both sides of the fight—and financial services organizations are caught in the middle of an arms race that’s accelerating by the quarter.
The FBI’s 2025 Internet Crime Report logged more than 1 million complaints and over $20 billion in reported losses, marking the highest figures the agency has ever recorded. That’s a staggering jump, but what financial services leaders should pay even closer attention to comes from the Association of Certified Fraud Examiners (ACFE) and Statistical Analysis System’s (SAS) 2026 Anti-Fraud Technology Benchmarking Report—a global survey of 713 anti-fraud professionals. Only 7% said their organizations are more than moderately prepared to detect or prevent AI-powered fraud, indicating that the other 93% are playing catch-up against criminals who’ve had a head start.
How this AI-on-AI battle plays out inside banks, insurance companies, and fintechs can mostly grouped into three buckets.
RELEVANT: Fraud Detection Still Rely on Human in the Loop Workflows
AI-Generated Phishing vs. AI-Powered Email Detection
Phishing has been around for decades, but generative AI has made it almost unrecognizable. Fraudsters are using large language models to craft emails that read exactly like internal communications, excluding previous tells like broken grammar, suspicious formatting, and “Dear Customer” openers.
These messages are hyper-personalized, often pulling details from LinkedIn profiles, press releases, and leaked data to mimic the tone and context of real business conversations. Business email compromise, one of the most damaging forms of fraud using AI, can now be executed at scale with fewer people and less effort than ever before.
These losses directly affect financial institutions. Business email compromise—one of the fraud types most supercharged by AI—accounted for $3.05 billion in reported losses in 2025, making it the second-costliest cybercrime category behind investment fraud, according to the FBI report.
The same report tracked 181,565 complaints with a cryptocurrency nexus and, for the first time, broke out AI-related complaints as their own category, logging $893 million in losses tied specifically to AI-facilitated schemes. Across the board, total reported losses due to AI jumped 26% year over year to $20.9 billion.
On the defense side, AI for fraud detection is evolving just as fast. Among organizations already using generative AI in their anti-fraud programs, 49% are applying it to phishing and scam detection, and 46% are using it for risk identification and assessment, according to the ACFE/SAS benchmarking report. Instead of scanning for typos and suspicious links, these systems analyze behavioral signals, communication metadata, and linguistic patterns to flag threats that would sail past a traditional rule-based filter.
Deepfake Impersonation vs. Biometric and Behavioral AI
What used to sound like science fiction is now posing real, everyday threats across all organizations, Fraudsters are using AI to clone voices, generate synthetic video, and bypass identity verification systems that were designed around the assumption that if you can see or hear someone, they’re real.
Voice clones are being used to impersonate executives on phone calls, authorize wire transfers, and defeat call center security protocols. Deepfake video is being deployed during remote onboarding and authentication processes to pass liveness checks.
The ACFE/SAS report found that deepfake social engineering saw the sharpest increase of any AI-powered fraud modality over the past two years, with 77% of respondents reporting a slight-to-significant rise. Consumer fraud and scams weren’t far behind at 75%, along with generative AI document fraud and forgery, also at 75%.
And it’s not slowing down. 55% of the anti-fraud professionals surveyed expect deepfake social engineering and GenAI document forgery to increase significantly over the next 24 months. In other words, organizations have 2 years to get their systems up to speed and vigilant.
Financial institutions are responding with AI-driven biometric defenses. Physical biometrics—fingerprint, facial recognition, and iris scanning—is now the most widely adopted emerging technology in anti-fraud programs, used by 45% of organizations, up from 34% in 2022, and a meaningful jump that reflects how urgently the industry is trying to close the gap.
Beyond physical biometrics, banks are layering in behavioral analytics that track how users type, navigate, and interact with devices over time to continuously verify identity rather than relying on a single checkpoint.
READ NEXT: The Identity Question Every Financial Organization Should Ask Themselves
Synthetic Identity Fraud vs. AI-Powered Anomaly Detection
Synthetic identity fraud has been a problem for years, but AI has supercharged it. Criminals use generative AI to build entire fake personas—complete with consistent names, addresses, government IDs, credit histories, and even social media profiles—that can pass onboarding checks and open accounts without raising red flags.
These fabricated identities behave like real customers for months, quietly building credibility before the fraudster maxes out credit lines, takes out loans, or executes payment fraud. By the time the institution detects the loss, the damage is usually already done.
An OmniWatch analysis of Federal Trade Comission Consumer Sentinel Network data found that identity theft reports filed between January and September 2025 already exceeded the total number filed in all of 2024. Credit card fraud—the single most common form of identity theft—saw its quarterly average spike 49.5%, with Q1-Q3 2025 reports reaching 503,450 and surpassing the full 2024 total of 449,090 by more than 12%.
Loan and lease fraud followed a similar trajectory, according to the same analysis, with Q1-Q3 2025 reports hitting 178,210 and matching 101% of the prior year’s full total in just nine months. These aren’t fringe categories. They’re the bread and butter of banking and financial services, and AI-generated synthetic identities are making them easier to execute and harder to catch at scale.
AI and fraud detection intersect most powerfully here. The the ACFE/SAS benchmarking report also found that a quarter of organizations now use AI and machine learning in their anti-fraud analytics, up from 18% in 2024, with another 28% planning to adopt within two years.
These systems can analyze millions of transactions and behavioral data points to spot anomalies that would be invisible to a human analyst—subtle inconsistencies in how a “customer” navigates a platform, timing patterns in transactions, or mismatches between stated behavior and actual activity.
However, systems can only do so much. That same report found that while 82% of organizations say explainability is an important factor in adopting AI for fraud detection, just 6% feel completely confident explaining how their models actually make decisions. For banks and insurers operating under strict regulatory scrutiny, governance is a serious liability.
Your Fraud Detection Is Only as Strong as the Team Behind It
AI can process data at scale, flag anomalies in real time, and adapt to new patterns faster than any manual review. But every model needs to be built, trained, governed, tested for bias, and explained to regulators. Budget and financial restrictions remain the number one barrier to implementing new anti-fraud technology, cited as a major or moderate challenge by 84% of organizations in the ACFE/SAS report. And while more than half expect their anti-fraud tech budgets to grow over the next two years, money alone won’t solve the problem.
Financial institutions need people who understand both the technology and the threat—fraud analysts who can work alongside AI systems, data engineers who can build and maintain detection models, and governance specialists who can ensure every AI-driven decision is auditable and explainable.
The AI arms race between fraudsters and fraud teams will keep escalating. Insight Global helps institutions pull ahead by investing in people running their defenses, just as much as the technology itself. Connect with us to start building your fraud team.
