Crypto crime stats show that such illicit activity represents a negligible amount of all cryptocurrency transactions — but current figures are merely the tip of the iceberg, and the true scale of crypto crime could be exponentially larger
Key insights:
-
-
-
Crypto crime is likely much bigger than it appears — Blockchain forensics firms only report what they can prove with 99%-plus accuracy, meaning the true scale of crypto crime is likely far larger than official reports suggest.
-
False negatives are still a problem — While achieving incredibly low false positive rates, these strict standards result in significant false negatives, with firms missing up to 75% of known criminal addresses in tested datasets.
-
This reporting gap reveals hidden losses — FBI data shows higher losses than do forensic reports and when accounting for the 85% of fraud victims who never report crimes, actual losses could exceed $110 billion annually.
-
-
Law enforcement has known about crypto-related crime for more than 14 years now. Five years ago, I felt these industry reports left a lot to be desired. A lot has happened in since then, however, and I have learned that clarity is becoming more important than caveats, because even my own caveated reports are being taken out of context by the cryptocurrency ATM industry.
The myth of “crypto crime”
Nick Furneaux points out “there is no such thing as cryptocrime” — spoiler: it’s all just financial crime. Yet, the blockchain forensics industry still has the annual tradition of issuing crypto crime reports that end up getting reviewed post hoc. However, my previous post showed how the prevailing reports appeared to prove Nick’s point, stating that crypto crime represented just 0.34% of all activity — effectively, a rounding error.
I wrote that these reports needed to be heavily caveated, as the figures identified were clearly smaller than the figures that may have been reasonably expected. In fairness to the industry, reports have since incorporated caveats on nearly all stated figures. However, this has still not stopped the industry from cherry picking figures that support the argument that there is no such thing as crypto crime.
The ironically good news in this year’s reports has been that the official figures for illicit activity across the industry has increased to more than 1% of all crypto activity for the first time since 2019. This increase is an indicator that the industry has gotten better at identifying criminal activity; and while there is still room for improvement, we are moving in the right direction.
Art vs. science
The companies producing these reports continue to hold some of the largest datasets on crypto-crime and blockchain metadata in the world. They are ideally placed to speak to these trends in illicit activity in the crypto ecosystem. However, one of the early arguments in blockchain forensics was that it is not as effective as some people were claiming.
In the landmark case, US v. Sterlingov (colloquially known as the Bitcoin Fog case), blockchain intelligence platform CipherTrace claimed that blockchain forensics was more of an art than a science. Based on evidence from Chainalysis, the case’s Daubert hearing acknowledged blockchain forensic evidence was admissible in criminal court to prove money laundering based on the methods used.
Understanding the limits of these reports requires an understanding of the core audience for these forensic firms: Law enforcement, which has a high burden of proof to achieve before going to court with any evidence.
Chainalysis has been doing this for 12 years at this stage and has been one of the only services to undergo a peer-reviewed academic study of its data, albeit a tiny sample size of its overall dataset. In the last five years, competitor TRM Labs has become an industry leader based on its focus on blockchain intelligence and law enforcement support.
The accuracy trap
Understanding the limits of these reports requires an understanding of the core audience for Chainalysis and TRM Labs: Law enforcement, which has a high burden of proof to achieve before going to court with any evidence. As such, the standard held by industry leading companies is that a data model should achieve an accuracy level of 99%-plus. However, as with any machine learning algorithm, it is incredibly difficult to guarantee 100% accuracy. Still, 99% accuracy is higher than human-based systems are expected to have.
Despite this commitment to high standards, the blockchain forensics industry has come under fire for false negatives. In the academic research of Chainalysis’ data, researchers found its false positive rate to be 0.01%, 0.15%, and 0.11%, respectively across the three datasets, or at least 99.85% accuracy for what was in their tool. Obviously, this is much more scalable and accurate in the modern world in which criminals are using AI than having humans unravelling these datasets manually. However, this level of certainty does paradoxically result in a surprising level of false negatives.
Indeed, Alison Jimenez, of Dynamic Securities Analytics, pointed out that Chainalysis missed a significant percentage of all addresses in the three sample datasets. The study looked at coverage of three known illicit services: BestMixer, Hansa Market, and Wall Street Market.
Chainalysis was found to have been able to identify 25%, 79%, and 95% of the sampled addresses, respectively. While this may seem like the company is negligent to suggest they can identify crime when it missed 75% of Best Mixer addresses, a service designed to obfuscate the flow of funds, the reality is that identifying any of these services is pretty difficult in the first place — especially in a world in which criminals are actively trying to escape surveillance. And remember, this is just the data that made it to production; Forensics firms are still able to assist law enforcement to make informed decisions on their investigations based on a range of additional data that never gets surfaced in the tool or in reports.
The reporting gap
These forensic companies are unable to publish informed estimates of the level of crime, but they are saying that they have identified at least $154 billion dollars in illicit activity in 2025. These tools also assist law enforcement with investigations which they may not always have permission to include in their datasets. Yet, investigators can still use the technology to carry out their investigations safe in the knowledge that their evidence will be admissible in court. That means, the $154 billion figure is effectively a floor, not a ceiling for the potential effectiveness of blockchain forensics.
The FBI counts what victims report, whereas forensic firms count what they can prove on-chain. When you consider that academic research suggests 85% of fraud victims never report their crimes to anyone, the scale of the problem becomes staggering.
The discrepancy between forensic reports and law enforcement data is where the caveats become most visible. The FBI’s IC3 report for 2024 (released in late 2025) pegged crypto-related scam losses at $16.6 billion. This figure is 67% higher than Chainalysis’s estimate, and 55% higher than TRM Lab’s for the same category.
Why the gap? Because the FBI counts what victims report, whereas forensic firms count what they can prove on-chain. When you consider that academic research suggests 85% of fraud victims never report their crimes to anyone, the scale of the problem becomes staggering. If we extrapolate the FBI’s reported figures to account for this silent 85%, the potential loss to crypto scams could be as high as $110 billion. While not an academically rigorous calculation, this figure would not surprise many industry analysts.
What will these reports look like in another 5 years?
The critique I have of these reports is that they underestimate the size of the problem in order to be able to accurately stand by their data. This isn’t a bad thing, it just results in unfortunate outcomes. There may be a day when these reports are combined with academic research to make a more informed estimation of how big the crypto crime problem really is.
Thankfully, those in the blockchain forensics industry can’t speak in theories or artistic interpretation. They have to be able to prove their statements and back them up with verifiable data. Right now, these reports are effectively looking at the tip of the iceberg and showing what they know about what they can see — the caveat now is that this is just the known knowns. The challenge continues to be identifying the known unknowns. Fortunately, we are getting better at identifying criminal activity every year.
You can find more of our coverage of the cryptocurrency industry here
