Europe Could Hold the Only Real Brake on the Global AI Race
A Guardian columnist argues that export controls on the machines that build AI chips might be the one lever powerful enough to slow an industry that few governments seem willing to touch.
Key points
- Columnist Alexander Hurst, writing for The Guardian, argues that the EU could use export restrictions on ASML, a Dutch company that makes the machines used to manufacture advanced computer chips, to force a slowdown in AI development.
- ASML holds a near-monopoly on extreme ultraviolet lithography machines, the equipment that prints the finest circuits onto silicon and without which cutting-edge AI chips cannot be made.
- Hurst lists automation-driven job losses, authoritarian surveillance, autonomous weapons and environmental damage among the risks he says the current AI boom is accelerating.
- He singles out tech figures including Elon Musk, OpenAI chief Sam Altman, Palantir co-founder Peter Thiel and venture capitalist Marc Andreessen as poor guides toward any positive outcome.
A Dutch company most people have never heard of makes the single most important piece of equipment in modern AI. Without machines built by ASML, chipmakers cannot print the tiny circuits that power the data centres behind ChatGPT, Google's Gemini, or anything else running at the frontier of artificial intelligence. That gives Europe a rare kind of leverage, and one writer thinks Brussels should use it.
Alexander Hurst, a columnist based in Paris, put the argument plainly in The Guardian: an EU ban on exporting ASML's lithography machines, the room-sized devices that etch circuits onto silicon wafers using beams of ultraviolet light, could be the circuit-breaker that slows an AI arms race no government has yet managed to stop by any other means.
What risks is Hurst actually worried about?
He names several, and they span a wide range. On the environmental side, he points to exploding carbon emissions from data centres. On the economic side, mass job losses through automation and the risk of a financial bubble bursting. On the political side, he raises the prospect of AI tools enabling authoritarian surveillance states and undermining democratic institutions.
The list goes further. Hurst flags the possibility of terrorists using AI to design bioweapons, of militaries deploying autonomous weapons systems that make lethal decisions faster than any human commander can follow, and of AI systems, conscious or not, pursuing goals that work against human interests. These are not fringe concerns. Versions of most appear in official risk assessments from governments in the US, UK and EU.
Who does Hurst blame?
He is pointed about this. Musk, Altman, Thiel and Andreessen, four men who between them shape enormous amounts of AI investment and policy thinking, come in for sharp criticism. Hurst argues that expecting this group to steer the technology toward environmental sustainability, social harmony or broadly shared prosperity requires a level of wishful thinking he compares to the kind of optimistic, uncritical output ChatGPT itself sometimes produces.
The column is opinion, not a policy proposal with a vote behind it. No EU institution has announced plans to tighten ASML export controls beyond restrictions already in place on sales to China. The Netherlands government has, under US pressure, already curbed some ASML exports since 2023, which shows the machine does function as a geopolitical tool when governments choose to pull it.
For ordinary readers, the practical question is simpler. The people building the most powerful AI systems are moving fast, and elected governments are mostly watching. Whether export controls on a single Dutch factory's output could genuinely slow that down is an open question. But Hurst's column, short as it is, puts a concrete mechanism on the table at a moment when most critics stop at describing the problem.
