The numbers hit me like a cold wave on a Tokyo morning.
Chainalysis's 2026 Crypto Crime Report dropped a figure that should make every one of us pause: approximately $17 billion lost to crypto scams in 2025. But the number that truly unsettled me is the multiplier. AI-powered scams extract an average of $3.2 million per incident — 4.5 times more than traditional schemes.
Criminals are not just using AI; they are weaponizing it daily. Voice cloning to impersonate a CEO. Deepfakes to bypass KYC checks. Automated phishing campaigns that target thousands of wallets simultaneously. This isn't a future scenario. It is the current operational baseline for a growing cohort of cybercriminals.
Meanwhile, law enforcement is struggling to keep pace. In some jurisdictions, investigators are literally forbidden from using AI tools. Others lack the training, the policy clarity, or the confidence to deploy the technology that criminals are exploiting every single day.
This is the AI crime gap. And it is the defining challenge facing our industry's next decade.
I have spent the last few years building a news team that prioritizes speed, but also understanding. When I decoded Compound's interest rate models during the 2020 DeFi Summer to explain the panic, I realized the same pattern: the human factor — fear, confusion, and a lack of clear information — is often the primary bottleneck. The AI crime gap is no different.
So, let's break down the technical reality, the market signals, and the contrarian opportunity that nobody is talking about yet.
The Criminal's Technical Advantage
We need to start with an uncomfortable truth. When it comes to AI, the bad guys are outperforming the good guys. Why? Because criminals are early adopters. They have zero bureaucratic overhead. No policy boards. No public accountability. They simply test, iterate, and scale.
Take the mechanics. AI doesn't just clone a voice. It can analyze a person's speech patterns, emotional tone, and conversational quirks to create a voice clone that is nearly indistinguishable. For the average retail investor, receiving a phone call from a "protocol founder" asking them to reset their wallet is a convincing trap. For a law enforcement officer, it's a nightmare to investigate.
Furthermore, AI scales. A human scammer might send 100 emails a day. AI generates 100,000 personalized emails, each referencing the target's recent on-chain activity. It automates the entire lifecycle: targeting, engagement, phishing, extraction, and laundering.
The Law Enforcement Bottleneck is Human, Not Technical
Here's the part that I believe gets under-reported. We have the tools to fight back.
Companies like Recoveris are building a high-confidence tracing of cross-chain, cross-bridge, and even mixer-obfuscated fund flows. The technology is there. So, why are we losing? I spent years auditing wallet addresses manually in 2017, and I know the frustration. But the challenge now is not the code. It's the bureaucracy.
Sol Cinosi, a former Buenos Aires prosecutor now at Recoveris, articulated this perfectly. He said the gap is not just about capability. It's about regulation. Some jurisdictions block the use of AI for investigations, fearing civil liberty violations. Others haven't updated their policies, leaving investigators in a grey zone of legal uncertainty.
Nick Pailthorpe, who spent 20 years in UK policing and now works with Kodex, hit the nail on the head. He explained that investigators are often afraid to use the tools they already have. They fear they don't have permission to use their own powers. They fear the public backlash. This psychological barrier is almost as strong as the technical barrier.
A silent epidemic of underutilization.
The Real Consequence: The 'Data Pile' Problem
We are drowning in data. Every transaction is a data point. Every bridge is a data cluster. Every smart contract is a logic gate. A human investigator can review maybe 100 transactions in a day, but the pattern of a sophisticated AI laundering scheme might be spread across 10,000 transactions. That's where the AI's value lies for law enforcement.
AI is not a magic bullet. It doesn't replace the detective. But it is the only tool capable of processing the sheer volume of data to identify the subtle, non-linear patterns that indicate criminal behavior. This is the 'Data Pile' problem. We have the 'Data Pile', but we are trying to dig through it with a teaspoon while the criminals are using an excavator.
The Contrarian Angle: The Rise of the 'RegTech' Economy
The mainstream crypto narrative focuses on the scam losses. That's the FUD. But let me offer a contrarian perspective that nobody is talking about yet.
This gap is a market signal. The AI crime wave is not just a problem. It is a massive business opportunity for the 'RegTech' (Regulatory Technology) sector.
Think about it. Recoveris is basically the 'Chainalysis for AI'. Kodex is the 'educational bridge' between exchanges and law enforcement. These are not just public-service tools. They are the next generation of infrastructure.
The demand side is exploding. Every major exchange, every custody provider, every DeFi protocol with a real security team needs to know how to detect AI-driven scams. The 'AI Investigators' are the new elite squad of the crypto world. They are the ones who will be able to sign off on a security audit and say, 'We are not just compliant with the letter of the law, but we can also survive the AI crime wave.'
The 'Law of the Horse'
There's a phrase in the tech world: 'the law of the horse.' It means you don't have a law of the horse; you have a law of transportation, property, and contracts that applies to the horse. In the same way, we don't need a 'law of AI'. We need to update the existing laws on fraud, theft, and privacy to apply to an AI-driven world.
That's where the real opportunity lies. Not in building more surveillance, but in building a more transparent and resilient financial ecosystem.
The adoption of crypto is growing faster than the number of experts who can analyze it. This is the fundamental contradiction. The Kodex model, which creates an education bridge between exchanges and law enforcement, is not a charity. It's a strategic necessity. If exchanges want to survive, they need to be proactive in educating law enforcement. The more they teach, the more they build a relationship of trust. The more they build trust, the more they can navigate the complex and evolving regulatory landscape.
The Takeaway: The Watchlist for the Next 6 Months
For me, the immediate takeaway is to watch the RegTech space. This is a 3-6 month window where the sector is likely to see more institutional interest.
Watch for policy changes. The moment the first major jurisdiction (I'm watching the UK and Hong Kong) announces a formal policy that explicitly allows AI use for fraud investigation, the dam breaks. The signal is not just about allowing the AI, but about approving the data-sharing agreements that come with it.
Watch the data. The $17 billion number is a baseline. If the 2026 report shows a decline in AI-scam average take, that's the signal that the tools are working. If it continues to increase, the gap is widening.
And watch the market. I am not saying we are about to see a 'Security Token' rally, but I am saying the security infrastructure is becoming a larger part of the crypto ecosystem's value proposition. The next bull run will not be driven by just the promise of money, but by the promise of security. The question is not whether we will catch up, but whether we can catch up before the criminals realize the next iteration of their attack.
Are we ready?
