Account Takeovers Surge as Banks Scramble to Respond

The imbalance underscores a broader shift. Criminal activity is pivoting toward external attackers regaining dominance after a brief period in which authorized-party fraud played a larger role. Banks are responding with higher spending and more advanced tools, but the contest remains uneven, as PYMNTS Intelligence and Block found in the “2025 State of Fraud and Financial Crime in the United States” report.

Unauthorized Attacks Regain Ground

Unauthorized-party fraud now accounts for most losses. These schemes, driven largely by compromised credentials and account takeovers, now represent 71% of all fraud incidents and total dollar losses. A year earlier, unauthorized parties accounted for less than half of reported incidents.

The shift reflects how quickly fraudsters adapt to controls and customer education efforts. Rather than relying on manipulation of legitimate users, attackers are increasingly exploiting stolen login information, altered payment instructions and direct account access to move funds at speed.

The data also showed that digital payment channels remain a prime target. Digital payment fraud accounts for the largest share of total dollar losses among unauthorized-party incidents, highlighting the exposure that comes with instant and always-on transaction environments.

Why Stopping Fraud Is Getting Harder

Even as banks improve detection capabilities, the environment around them is becoming more difficult to manage. Nearly half of financial institutions reported that fraud schemes have become more sophisticated over the past year, up from 35% in 2024. At the same time, institutions must contend with faster payment speeds, more payment types and growing regulatory demands.

These pressures especially collide at large banks, which operate complex legacy infrastructures while supporting real-time or near-real-time payment experiences at scale. The result is a fragile equilibrium in which institutions blunt some attacks but struggle to get ahead of the next wave.

Top Fraud Challenges Facing Financial Institutions

According to the report, banks and other financial institutions cited the following as their most pressing fraud-related challenges:

• Increasing sophistication of fraud schemes (46%)
  Criminals are using more advanced techniques that evade traditional, rules-based controls.
• Cost of reimbursing fraud-related losses (46%)
  Direct financial exposure remains a burden, particularly for large banks.
• Financial sanctions compliance (47%)
  Expanding sanctions regimes add complexity to transaction monitoring and enforcement.
• Increased speed of payments (46%)
  Faster settlement leaves less time to detect and stop suspicious activity before funds move.
• Expansion of payment types and currencies (41%)
  New rails and formats create additional attack surfaces for fraudsters.
• System integration limitations (35%)
  Legacy platforms struggle to connect with modern fraud detection tools in real time.

How Banks Are Defending Themselves

Despite the challenges, most institutions are not standing still. More than two-thirds of financial institutions increased fraud detection spending year over year, signaling that security has become core infrastructure rather than a discretionary investment.

Machine learning and behavioral analytics are now foundational elements of modern fraud defense. Roughly 8 in 10 large banks reported using advanced behavioral analytics, and three-quarters have implemented machine learning or AI models to support proactive detection and reactive response.

At the same time, banks are pursuing layered strategies. About half plan to expand cloud-based fraud platforms or outsource certain detection functions, while others are building new in-house systems or improving communication with customers to reduce losses and restore trust.

What emerges from the data is not a single solution but an ongoing balancing act. Fraud prevention is increasingly defined by continuous adaptation, where banks must modernize defenses while managing speed, scale and regulatory complexity in parallel.