Last week, with barely a whimper of protest, Britain was cut off from the most powerful technology on the planet – and consigned, I fear, to a future as a defenceless, third-rate power.

The two most powerful AI models in the world – Mythos 5 and Fable 5, both created by the San Francisco lab Anthropic – were slapped with ‘export controls’ by the US government and closed to the outside world.

The ban appears to have been triggered by the discovery that the models, in the wrong hands, could become the ultimate hacker’s tool – even, it is rumoured, exposing weaknesses in the software systems of America’s National Security Agency.

In one of the last acts of his lame duck premiership, Sir Keir Starmer made a token attempt to resist, pleading with Donald Trump for Britain to retain access, only to be told by the White House: ‘Zero chance!’

Not all readers may be aware of this historic snub and its potentially catastrophic implications.

The political circus within the Labour Party and the shuffling into post of yet another prime minister are of vanishingly small importance compared to the existential consequences of being locked out of this technology.

Fable and Mythos are AIs that code; writing and editing software to levels far beyond human capability. This makes their hacking abilities more powerful than models such as ChatGPT, whose parent company OpenAI has focused its latest efforts on image and video creation. Fable and Mythos are now so brilliant at coding that they could hack into most of the world’s banking, healthcare and security systems with ease.

I’ve worked in both public policy and at an AI company in London. In both roles, I tried to warn about the importance of preparing ourselves to face the transformational onslaught of this technology – often in vain.

Industry bosses and even governments refused to listen. So last year I quit my job to make a documentary, out this year, about the threat AI poses to life as we know it.

The Anthropic ban is so strict that even former prime minister Rishi Sunak – a senior adviser to the $965 billion company (the world’s largest firm focusing purely on AI) – is believed to be barred from accessing the software.

Before, it was taken for granted that any friendly country, particularly Nato allies, could benefit from American technology if they were willing to pay for it. Now we have to rely on the whims of the sitting US President.

AI has already been used in the pinpoint bombing that destroyed Iran’s air defences and killed its leaders last February, as well as, reportedly, to plan the surgical operation that removed Venezuelan dictator Nicolas Maduro two months earlier.

AI is also revolutionising drone warfare, which has become infinitely more sophisticated since it emerged after the Russian invasion of Ukraine four years ago. Swarms of AI-controlled drones, whether operated by America, China or Russia are on the horizon.

This week one Nato chief, Major General Adrian Ciolponea, warned that the alliance is having to prepare for the prospect of this new warfare. ‘Nations without this type of [swarm] technology will be forced to join a security organisation or to accept the conditions of an aggressor,’ he said.

Currently, the best drone operators are video-gamers, mostly Gen Z whizzkids who grew up on PlayStation and Xbox. But even the most deadly and proficient cannot think or react at AI speeds – nor can they control swarms of hundreds, even thousands, of drones simultaneously.

Once AI is able to control unmanned aerial weapons better than humans, the only limit to a drone assault will be the capacity of war factories to produce the drones themselves. They could literally be as numerous as locusts, the skies black with them, each hunting a different target through facial recognition.

To be cut off from these advances is devastating. The prospect has left informed policymakers reeling – even if most British politicians appear shamefully ignorant of the problem.

The US could even exploit its edge in AI as leverage over Europe. By withholding access to certain models, America could force countries into mind-boggling concessions – such as pressuring Denmark into ceding control of Greenland.

As French presidential candidate Bruno Retailleau warned: ‘In the race for artificial intelligence, a nation that depends on others for its technology is a nation that can be unplugged overnight.’

Former Tory security minister Tom Tugendhat, who does understand these issues, put it even more simply: ‘We cannot continue like this and remain sovereign.’

There is a historical precedent for this crisis.

Eighty years ago, when the Second World War ended with the destruction of Hiroshima and Nagasaki, the US refused to share its nuclear secrets with its wartime allies. Britain was forced to go it alone, developing our own H-bomb by 1952, though it was four more years before the weapon was operational.

That left us without an independent nuclear deterrent for a full decade at the start of the Cold War – a particularly perilous time.

Crucially, it was not until we proved we were capable of defending ourselves that the US belatedly agreed to share much of its technical intelligence – a partnership that has lasted ever since.

The current government appears entirely ignorant of these issues. As if to add insult to injury, Labour’s approach towards ‘Big Tech’ – which is responsible for almost all the West’s economic growth, with seven tech giants now making up fully 70 per cent of the US stock market – is one of ignorant suspicion.

Starmer’s administration has seemed intent on trying to regulate the industry – by which it largely means Facebook owner Meta, rather than the pioneering labs driving most of the innovation – out of existence.

National security considerations aside, Trump separately nurses a long-held grudge against Anthropic and its liberal, anti-Maga CEO Dario Amodei. It’s entirely possible that the President, who once called Anthropic ‘woke Lefties’, was partly acting out of spite when he issued the ban.

Whatever his motive, he is hurting America’s friends more than its enemies. China has spies in every major tech giant and flaunts its readiness to steal Western AI secrets and intellectual property. Beijing’s digital ‘innovations’ are typically just six months or so behind the latest from US developers.

As long as Britain is barred from these technologies, all of us will be less safe.

Days after Trump announced the Anthropic ban, a group of British, Dutch and German researchers issued a paper called ‘Europe 2031’. In their hypothetical but realistic timeline, China and the US will vie for AI supremacy, leaving Europe so far behind that by 2029 our continent will be on the brink of bankruptcy and begging America for a bailout.

We have to wake up. Whoever becomes PM must rethink the defence agenda in light of this creeping catastrophe. Britain cannot rely on the US to share this technology. Yes, it will require significant investment but this is a defining moment for our national security. We must work together with other powers – such as Canada, France, Germany and India – to give our Armed Forces the best AI systems in the world.

Or we will pay the price – sooner than anyone thinks.

Connor Axiotes is head of Tail End Films whose documentary Making God, an exposé on the race for artificial general intelligence, is released this year