AI Scam Hits $3.2 Million Per Incident, Crypto Security Battlefield Shifts from 'Code' to 'Scams'

On-chain scams reach $14 billion a year: AI has raised the average single incident scale to four times that of traditional scams.

Written by: Boaz Sobrado

Compiled by: AididiaoJP, Foresight News

Identity impersonation and AI-driven scams are becoming one of the biggest security threats in the crypto space. According to Chainalysis data, on-chain funds flowing into crypto scams are expected to reach at least $14 billion by 2025. Of course, not all of this is related to AI. Rime Salmi, founder and CEO of Fractl, said in an interview: "Many people pretend to be investors, but they are not. They are looking for investments or are simply scamming." She believes that technology itself is blurring the boundaries: "AI is a buzzword, but we can no longer tell where the person is. If a company has only one person, can we still accept that? Or do we still only invest in solid large teams?"

The Economics of AI Fraud

Salmi stated that AI is no longer a standalone entity: "It is embedded in everything. So it is no longer an independent topic." Chainalysis statistics show that scams associated with AI vendors have an average operational scale of about $3.2 million; scams without such associations average about $719,000. The company does not express this correlation as a causal relationship. Salmi believes that the current defensible line is data that cannot be copied by others: "If a startup has solid proprietary data, it is difficult for others to push it out with 'ambient programming' because truly high-quality, in-depth data takes years to accumulate."

In this subset identified by Chainalysis, deep fakes, face-swapping software, and large language models are used to create fake identities on a large scale. Erika Maslauskaite, CEO and co-founder of AlongID, said in an interview: "With today's AI, you can almost fake anything. Really, anything." She summarized the issue as: "How do we verify what is real and what is fake?"

Hacker Techniques are Evolving

AI scams have not rendered traditional hackers obsolete. Attackers do not need to breach smart contracts; they only need to take down the person authorized to use the contract.

Jimmy Su, Chief Security Officer of Binance, stated: "Code is no longer necessarily the weakest link in Web3. As smart contract security improves, attackers are turning their attention to the people, credentials, and governance systems surrounding the protocol. We have experienced this firsthand: the Binance security team assisted in stopping a $1.2 million governance attack on BrainTrust. Today, protecting a protocol is not just about protecting the code, but also about protecting who can control it, how control is exercised, and the underlying infrastructure and people behind it."

In April 2026 alone, access control failures accounted for about two-thirds of the $621 million in DeFi vulnerability losses, according to data from Binance Research. Nitya Subramanian, founder and CEO of Para, said on the On The Margin podcast: "Wallets are essentially a layer of authorization and control flow for everything on-chain. Every chain, every DeFi primitive, every on-chain operation has to go through a wallet. I think many people still do not truly understand this."

Attribution Challenges

Blockchain transparency remains one of the strongest evidentiary advantages in crypto, but attribution requires context beyond transactions. Dmitry Machikhin, founder and CEO of BitOK, pointed out in an interview that mixing coins remains a gap: "Even Chainalysis does not have a 100% solution for mixing coins." He said that seizure notices do not always come to fruition: "There was a case where the Israeli government seized certain wallets in official documents. Logically, any exchange or entity dealing with crypto should block these wallets. But that did not happen."

Machikhin stated bluntly: "We have not caught anyone. We are just showing the path of funds." He estimated the volume of illegal transactions to be "less than 0.1% of all transactions," and then added: this proportion underestimates the problem, "even 0.1% is already significant, and it is still growing with the market." He also pointed out that crypto still dominates a certain channel: "In terms of terrorist financing, fiat currency is actually more efficient. But in the dark web, crypto remains the first priority path."

The Chainalysis 2026 report shows that UK law enforcement recovered over 61,000 bitcoins in 2025 and seized $15 billion related to Prince Group.

Next Target May Not Be Humans

Varun Kabra, Chief Growth Officer of Concordium, said on the On The Margin podcast: "The next step has already begun: AI agents are starting to trade for you. They pay, register services, and are likely already handling your financial transactions."

The gap is at the receiving end: "The counterpart on the other side—like airlines, ticketing platforms—cannot verify whether there is a accountable real person behind this transaction. This opens a door to fraud: robots impersonate humans, agents operate without any accountability."

Kabra predicts that agent traffic may surpass real human transactions in the next 6 to 12 months and states, "The accountability between humans and agents is, in my opinion, the biggest problem that needs to be solved in this world." Subramanian describes the same change from the user side: "Agents essentially outsource purchases. Anyone who has outsourced procurement knows that this brings trade-offs."

Concordium's Agent Registry provides agents with on-chain identities linked to verified real owners. The company stated in July that as of July 14, 2026, there were 1,131 agents on the registry, with over 15,663 on-chain transactions. Kabra emphasized that this binding does not equate to exposure: "I have always believed that privacy and anonymity are two different things." The mechanism he describes is "selective disclosure, with zero-knowledge proofs, so no one knows it is you."

Atul Khekade, co-founder of XDC Network, said in an interview that the track has not yet been laid: "AI currently does not have a transaction layer." The compliance layer is also lacking: "AI platforms do not have a monetized compliance layer that can be used for real transactions and to execute actions." The counterpart will not quickly fill this gap, "expecting them to build this infrastructure overnight is unrealistic."

Defense is Also Being Automated

Chainalysis states that its Alterya platform can help banks and crypto companies identify known scam addresses before funds exit. Machikhin was straightforward about this: "We cannot survive without AI, right?" He said his tracking also relies on similar tools, "with our intuitive AI product, it can mark the block paths, and we can follow."

Maslauskaite referred to the gap as "the missing trust layer on the internet." The raw materials are scattered everywhere: "Our digital identity attributes are all over the place, and we cannot control them at all." Crypto can verify the transaction itself but cannot verify the assumption that the person authorized to do it. Regarding agents, she drew the same line as Kabra: "Every agent must verify who is acting behind them."

What Will Happen Next

Khekade stated that demand has outpaced infrastructure, and the market is "growing almost at the speed of light." Criminals do not even necessarily need code vulnerabilities. Maslauskaite also believes that the rules are lagging behind: "The speed of technological development far exceeds the regulatory knowledge we have."