Warnings that China must be cut off from advanced AI chips echo a familiar pattern. History suggests technology bans rarely slow China down – and often do the opposite.

New year, same China threat, more tech‑savvied.

The latest comes courtesy of Dario Amodei, CEO of Anthropic, as reported by News Corp, warning that China’s access to AI chips has to be banned before it’s too late.

Selling advanced AI chips to China, Amodei tells us, is “like selling nuclear weapons to North Korea”.

You can almost hear the music swell, as if Dr Fu Manchu has just ordered the Death Star to lock onto the White House.

The only solution, according to the report, is the chips.

Chips, according to Amodei, are “the single greatest bottleneck” to powerful AI. Block them, and the threat recedes. Don’t block them, and we risk an “AI totalitarian state” and even military conquest.

On paper, the story hangs together neatly.

Chips are dangerous. China is dangerous. Therefore, chips plus China is your catastrophe combo, ready to serve to the public .

And so, the argument goes, sanctions are not just prudent policy. They are a moral obligation.

But let’s stop there.

The first problem isn’t subtle. It’s structural.

AI chips sit much closer to engines than warheads — tools that keep hospitals running, supply chains moving, forecasts improving, and research advancing. Like steel, electricity, or satellite imagery, they carry obvious civilian uses alongside military ones.

Labelling them as nuclear weapons does the work of persuasion without the burden of explanation.

A complicated, dual‑use technology is reduced to a single moral lever: export becomes betrayal, restraint becomes virtue.

Once framed that way, the argument largely runs itself. Evidence becomes optional.

Nowhere in the reporting do readers hear that AI development doesn’t sit inside one sealed bunker waiting for permission slips. It spreads. It adapts. It leaks. It improves under pressure.

But “diffusion” doesn’t sound as good as “doomsday”, especially when there’s another detail that rarely survives the edit.

Before warning Washington about the dangers of Chinese AI, Amodei worked at Baidu – China’s leading technology company – on its AI team in 2014.

He then moved to Google, then OpenAI, and now occupies the role of trusted US defence contractor, with a $200 million Pentagon framework deal signed in 2025.

Media coverage treats China’s response to sanctions as if it hasn’t happened yet – as though we’re still waiting to see whether pressure works.

In reality, we’ve been watching the response unfold in real time.

Instead of begging for access, Chinese companies have started handing things out, producing domestically designed accelerators at roughly 40 per cent of Nvidia’s performance, with the gap shrinking fast.

Not only products. But also blueprints.

While American firms keep their AI systems behind paywalls and usage rules, Chinese companies are going open‑source, making everything available to the world, publishing the full cookbooks – giving away the recipes so others can bake their own.

It’s a different business model.

And it has obvious appeal to countries that don’t want to rent their future from Silicon Valley.

The media rarely frames it that way. “Open source” sounds technical. “Free tools” sounds political. “China generosity” sounds suspicious.

So the story reverts to safer ground: China threat, western restraint, inevitable conflict.

The most striking omission in this entire debate is history.

Western technology exclusion has a remarkably consistent track record when it comes to China.

China was pushed out of Europe’s Galileo satellite navigation project. It responded by building BeiDou – now more precise, operating more satellites than GPS, and used widely across the developing world.

Barred from the International Space Station, China built Tiangong instead – larger, smarter, and equipped with ion propulsion. Their taikonauts now enjoy barbecue chicken wings in orbit, while ours still squeeze toothpaste meals from tubes.

And next? Crewed lunar missions before 2030. Not bad for a country we tried to keep out of space.

Each time, the same assumption was made: exclusion would slow progress.

Each time, the opposite happened.

That’s the part we keep erasing.

And there is one detail so awkward it rarely appears in these stories at all.

In 2013 – long before today’s panic about killer robots – China, together with 30 other countries, proposed a global ban on fully autonomous weapons.

The proposal went nowhere.

It was rejected by the United States (and many of its allies).

Fast‑forward a decade, and we are told China might use AI for military domination – a scenario the US explicitly declined to outlaw when given the chance.

Do we really need to sell chips to China – or nukes to North Korea – to destroy the world?

Or, shall we look at the facts?

Western technology exclusion has a remarkably consistent track record when applied to China – at least in domains where knowledge is already relatively diffuse and substitutes are technically feasible.

BeiDou overtook Galileo in both coverage and precision after China was pushed out of the project.

Tiangong became operational while China remained barred from the International Space Station.

Huawei, despite being cut off from leading-edge foundries, still built a globally scaled 5G ecosystem.

This pattern doesn’t guarantee anything.

Frontier AI – especially compute-intensive foundation models – may prove more bottlenecked than previous domains.

But if history is any guide, exclusion rarely delivers the clear, permanent gap policymakers hope for.

What it often delivers instead is urgency, acceleration, alternatives, so one country which wants to make itself great again ended up propelling its rival.