AI panic from a crypto veteran: why the race could shake banks and crypto
Crypto pioneer warns unchecked AI could strain banking systems and expose crypto firms, highlighting a competitive race with few safeguards.
Lead
Marc van der Chijs, co‑founder of former bitcoin miner Hut 8, told CoinDesk he’s become a "doomer" about artificial intelligence. He argues that the unchecked speed of AI development could strain legacy banking software and leave crypto businesses more exposed than the blockchain itself.
What happened
In a September 17 interview, van der Chijs warned that “we have lost control” of AI’s trajectory and that the pace may outstrip any emerging international safeguards. He likened today’s AI race to the early days of bitcoin in 2013, when a small group of enthusiasts believed they were shaping a new financial system. Today, he sees a structural trap: companies and nation‑states racing to deploy ever‑more capable models, penalising any restraint because competitors can capture market share or strategic advantage.
The interview also revealed his personal shift. After selling much of his bitcoin, he poured capital into AI ventures, calling Hut 8’s pivot to AI data centers “the best move ever.” He believes AI and robotics could soon perform 90‑95 % of existing jobs, driving down the cost of goods and services and forcing governments to find new revenue streams, perhaps through taxes on robots or AI token usage. While he remains bullish on AI’s economic impact, he cautions that a major disruption—potentially in finance or critical infrastructure—may be required before governments agree on guardrails.
Why it works that way
The core of van der Chijs’s concern is the feedback loop inherent in frontier AI development. When a model can improve its own code—a process known as recursive self‑improvement—each iteration can become markedly more capable without proportional human oversight. This accelerates progress faster than traditional R&D cycles and creates a competitive advantage for early adopters. Companies race to capture that advantage because the market rewards speed with higher valuations, while nation‑states chase strategic superiority, especially in defense and economic power.
Legacy banking software is built on decades‑old codebases, often written in languages and architectures that lack the flexibility to integrate rapidly evolving AI tools. When AI systems are woven into transaction processing, fraud detection, or risk modelling, hidden dependencies can emerge. A flaw in an AI model—whether a data bias, a mis‑aligned objective, or an unexpected interaction with older code—could cascade through the tightly interconnected banking network, eroding confidence in institutions that rely on smooth, real‑time settlement.
Within crypto, the network layer (the blockchain) is designed to be permissionless and resilient: every node validates transactions independently. Exchanges and other services, however, sit on top of that layer and rely on traditional infrastructure—servers, APIs, third‑party data feeds. If AI tools that power market‑making, order routing, or compliance malfunction, the resulting outage or manipulation could be far more damaging to the business than a protocol‑level attack. Van der Chijs argues that those businesses are more vulnerable than bitcoin itself because they cannot fall back on the decentralized consensus that protects the underlying ledger.
What changes because of it
Van der Chijs’s warning does not predict a specific crash, but it highlights three practical shifts for investors and regulators. First, capital is already moving from crypto mining toward AI data centres, a trend he believes will continue. That reallocation could keep bitcoin prices below the $200,000‑$250,000 range he once expected, as mining profitability wanes and investors chase higher‑growth AI opportunities.
Second, the risk profile of crypto businesses may rise. As AI becomes a core part of exchange operations, auditors and regulators may demand proof that AI models are auditable, that they can be rolled back, and that they do not create systemic points of failure. Exchanges that can demonstrate robust AI governance may gain a competitive edge, while those that rely on opaque, third‑party models could face heightened scrutiny or loss of user trust.
Third, the broader financial system may see a push for stronger guardrails. If a high‑profile AI‑induced outage hits a major bank or a crypto exchange, lawmakers could accelerate the creation of international standards for AI safety, much like they are currently discussing for stablecoins in the APAC region. However, as van der Chijs notes, the market may need a “shock” before coordination emerges, because the current race rewards speed over caution.
In practice, this usually means investors should watch for two signals: a rise in AI‑related governance disclosures from crypto firms, and any regulatory proposals that tie AI model transparency to licensing for financial services. Those who understand that the underlying blockchain remains technically robust, while the surrounding infrastructure becomes a new attack surface, will be better positioned to allocate capital.
