Fraud losses are up 430% since 2020. The window to stop them has not grown at all.

Key takeaway

Reported consumer fraud losses reached $15.9 billion in 2025, up nearly 430% since 2020. The biggest losses come from payments victims are persuaded to authorize themselves, and the money moves fast. Detection keeps improving, yet losses keep compounding, because the decision comes too late. Stopping fraud means deciding inside the authorization window, with state, decision logic, and enforcement in a single transaction boundary.

Earlier this year, the Federal Trade Commission put a number on what every fraud team already knows. Testifying before the US Congress Joint Economic Committee in March, the FTC reported that consumers filed 3 million fraud reports in 2025, with $15.9 billion in reported losses, up from over $12 billion in 2024.¹ Reported losses have risen nearly 430% since 2020.¹ And because most victims never report, the FTC estimates the true cost of fraud in 2024 alone may have been as high as $195.9 billion.¹

Behind the headline number are three details that matter more than the total.

The growth is driven by large losses. The FTC attributes the trend to a sharp rise in consumers reporting losses of $100,000 or more, and investment scams alone account for roughly half of all reported losses. Bank payments carried the highest aggregate losses of any payment method, ahead of cryptocurrency.¹ In other words, the biggest losses come from payments victims are persuaded to authorize themselves.

The money moves fast. PYMNTS Intelligence research found that nearly two thirds of scam victims send money within 24 hours of first contact.² By the time a case is opened, the funds are gone.

Impersonation is the delivery mechanism. Imposter scams were the most reported fraud category in 2025, with more than 1 million reports to the FTC,¹ and the same PYMNTS Intelligence report also found that more than 8 in 10 reported scams involve impersonation.² Generative AI has made convincing impersonation cheap and scalable.

Detection is not prevention

The industry has spent a decade getting better at detecting fraud. Scores are sharper, models are richer, and vendor signals are everywhere. Yet losses keep compounding, because in payments there is exactly one moment when detection is worth anything: the window between transaction submission and settlement. Once the money moves, especially on instant rails, recovery is slow, incomplete, and expensive. On FedNow, RTP, and similar rails, it is often impossible.

Most fraud architectures cannot act inside that window because they separate the three things that must happen together. The state needed to score accurately sits in one system. The scoring logic runs somewhere else, often behind an API call. The enforcement instruction travels through a queue to a third system. Each hop adds latency and, worse, staleness. A fraudster firing rapid parallel transactions exploits exactly that gap: each check reads velocity counters that do not yet reflect the others. Dheeraj Remella, Field CTO at Volt Active Data, calls this the systems gap.

That is why the answer to a 430% rise in losses is not a faster model. It is an architecture where state, decision logic, and enforcement share a single transaction boundary, with the decision completed and recorded while the authorization is still open.

What does decisioning in the window require?

• Authoritative state at decision time. Velocity counters, device history, behavioral baselines, and account balance read atomically, with no stale views under concurrent load.
• Scoring co-located with state. Rules executed against that state in the same transaction, with ML scores consumed as inputs alongside it, not stitched together across external API hops.
• A decision the rail can enforce. The approve, decline, or step-up decision is recorded as authoritative and returned inside the authorization window of the rail, whether that is a card network or an instant payment scheme.
• Rule changes at fraud speed. Scam patterns now shift in hours. Fraud teams need to deploy new logic multiple times a day, atomically, under live load, without a throughput penalty.
• A complete audit trail. Every decision recorded with the state it was made against, for compliance, disputes, and model retraining.

This is the approach behind results like the 83% reduction in fraud inside the authorization window at one institution using Volt as its real-time decisioning layer, with rule logic hot-swapped under live production load.

The AI question is coming next

The next phase of this arms race is agentic: AI systems that investigate suspicious activity, reason across signals, and recommend responses. That raises the stakes on data quality, because an agent reasoning from stale state produces confident recommendations built on the wrong facts. Dheeraj argues that AI should do the heavy lifting on the hard cases while deterministic, auditable logic keeps authority over the final decision. Most operational decisions are routine and should be handled deterministically. Volt makes those decisions in real time, against authoritative state, and escalates only the ambiguous or novel cases to ML models, AI agents, and human reviewers. Agents earn their place on the novel cases, where they need to query live, authoritative state as they reason, and where every query, recommendation, and final decision is recorded for audit and retraining.

The bottom line

The FTC data confirms the direction of travel: bigger losses, faster movement, and rails where reversal is not an option. Institutions that can only detect fraud will keep absorbing it. Institutions that can decide inside the authorization window can stop it before it becomes a loss.

Volt Active Data is the real-time decisioning layer for fraud prevention, real-time payment rails, and payment gateway infrastructure.

Sources

1. Federal Trade Commission, prepared statement, “The Rising Scam Economy: Modernizing Federal Approaches to Protect Americans from Foreign Fraudsters”, US Congress Joint Economic Committee, March 25, 2026.
2. PYMNTS Intelligence, “Financial Scams and Consumer Trust”, commissioned by Block, November 2025.