The spectre of AI fakes on platforms like TikTok looms large over the general election - David Behrens

Seldom has there been a campaign more negative than this: an exercise in nothing more than finding the least worst option. The arrival of Nigel Farage on the campaign trail demonstrated this better than anything, for here is a man whose sole pitch to voters is to vanquish his former fairweather friends in the Conservative party. He doesn’t seek to build; just to destroy.

Farage is a deepfake in human form but as we enter the last four weeks of electioneering – yes, there’s nearly a month more of this – the hustings are awash with even more insidious falsehoods. And unlike Farage’s manifesto, some of them may actually fool you.

AI has been getting a bad press but that’s partly because most of us don’t fully know what it is. So let’s set the record straight. Computers have always been good at running calculations on data they’ve been given – the numbers in your tax return, for instance. But AI allows them to make computations based on information they have only inferred. That’s a subtle but important difference.

And without getting too technical, once an AI program has determined something to be true it can go on to make further assumptions using the same logic – a process known as generative AI.

Like most technology, that can be a force for good or ill, depending on who is pressing the buttons. The NHS has used AI to identify tiny signs of breast cancer which had been missed by human doctors. Less helpfully, Poland said this week that a fake news report on mobilising 200,000 men was likely the work of a Russian AI robot.

But you don’t need to work in the Kremlin to deploy deepfakery; anyone with a phone or computer can do it, with no money and almost no skill. I proved the point by trying it myself. I used one of the many voice conversion sites on the web to record myself promising to build 800 new houses every minute during my term in office and converted my voice to that of Donald Trump. I’d have used Keir Starmer’s voice but the website didn’t know who he was.

Of course, no-one other than me heard any of this but the purpose of the site was clear: to share its creations on Facebook, TikTok and every other propagator in a worldwide game of Chinese whispers. It’s exactly how lies are started.

And most of those untruths are far more insidious than mine. A BBC report this week found that TikTok was actively feeding young voters in marginal constituencies fake clips of party leaders and soundbites littered with abusive comments.

This is worrying because a lot of young people are not too discerning about where they get their news. According to the media regulator Ofcom, around one in 10 people now cite TikTok as their primary source of information and the figure is growing year on year.

TikTok and AI have both emerged since the last election and politicians lack the tech skills to deal with them. For heaven’s sake, it’s only been five years since Boris Johnson manoeuvred himself into a fridge on live TV while shouting ‘Get Brexit Done’.

The reason TikTok has succeeded is clear: like all social media sites it presents readers with information they want to hear, feeding their prejudices and validating their beliefs.

That’s not a new science. Polarising newspapers like the Daily Mail and The Guardian do the same, but they are circumspect in what they print. TikTok has no such filter; anyone can say literally anything and they do. And if one person in 10 takes it as fact, the truth is fighting a losing battle.

In the US they have learned this already. The aftermath of Donald Trump’s trial saw people finding their own ‘truth’, casting him as the victim of a liberal conspiracy rather than the corrupt and seedy adulterer the jury found him to be. Boris Johnson was unfortunately right – for once – when he observed that the verdict would help, not hinder, his path back to the White House.

Yet the US is ahead of us in implementing root-and-branch measures to control misinformation. Last month it signed into law a bill that will ban TikTok across America unless its Chinese owner sells it off. If you’re going to stop AI regenerating itself more often than Doctor Who, that’s the only way to do it.

On the other hand, the logical result of letting generative AI run its course is a computer clever enough to be elected prime minister. Would that be a step forward or back?