In the first half of this year crypto startups received $10bn in venture capital through nearly 750 deals. VENTURE CAPITALISTS are pouring so much money into artificial intelligence that little is left over for anything else. Besides using lots of computing power, AI and crypto have had little in common. But an investigation by The Economist suggests the two are tightly linked by a booming illicit trade.
Some, though, retain a soft spot for an earlier mania: cryptocurrency finance.
Issuance of crypto-enabled credit and debit cards is surging. These cards connect crypto wallets to payment terminals used by ordinary businesses. They allow people in poor countries with unreliable currencies to transact in stablecoins pegged to less flaky foreign monies, notably the dollar. But crypto cards have other big customers, too. Many card providers do not make serious attempts to verify the identity of their cardholders. That has left them open to abuse by criminals and sanctions evaders. They also let people in places like China buy subscriptions to American frontier AI models, circumventing the model-builders’ restrictions on sales to America’s adversaries. Crypto cards convert digital currencies into real-world money when swiped at Visa and Mastercard terminals, and their use is growing. ether and other unstable coins, the volatility of unpegged cryptocurrencies makes them a poor means of exchange Although such cards can be linked to bitcoin. The true figure is probably higher; many crypto-card providers do not report their transaction volume. The spread of stablecoins worries regulators in poor countries, since it could lead to pseudo-dollarisation of their economies. But it should concern their rich-world peers, too.
Stablecoins, whose overall market value has nearly doubled in the past two years to over $300bn, are still mostly stores of value. But stablecoin cards were used in transactions worth more than $1.1bn in August, three times as much as a year earlier, according to Paymentscan, a data provider. Illicit crypto flows surpassed $158bn last year, estimates TRM Labs, an analytics firm.
Stablecoins, notes Eswar Prasad of Cornell University, “create additional conduits for illicit financial transactions, both within and between countries”.
In one instance, his firm observed a user with a Shenzhen IP address and China-issued ID apply for a card with a non-Chinese address; the user then spent $50,000 on ChatGPT subscriptions in the Bolivian Apple app store. Anthropic alleges that to do so Alibaba used a pool of 5,000 Claude accounts, some paid for with virtual cards to hide its actions. For businesses that want to block certain customers, crypto cards can make it difficult to verify a cardholder’s identity or location. Persian- and Russian-language sites, for example, advertise their own or third-party crypto cards to, among others, Iranians and Russians, who are locked out of the Western financial system by sanctions against their countries. A report published in April by Crystal Intelligence, another analytics firm, found that nearly seven in ten crypto-card providers had weak KYC requirements or none at all. About half of the no-KYC providers identified by Crystal Intelligence were based in Hong Kong and a fifth in America. Others sell virtual cards, in some cases entirely legally, while being licensed in the Caucuses or Gulf countries, where KYC rules are weaker and less vigorously enforced. Alibaba did not respond to a request for comment.
One Russian-language site boasts that they enable users to “pay for ChatGPT, Claude, Netflix, Spotify…and other international services”. A source at an American crypto provider whose cards are issued by Rain says that each month he sees hundreds of users signing up for his firm’s Visa cards with China-issued IDs with addresses outside China. Rain says it is investigating these reports. On September 10th Anthropic accused Alibaba, a Chinese internet giant, of illicitly gaining access to the Claude model 151m times between May and July to train its own models in a process known as “distillation”.
