On Thursday the 27^th of June, Channel 4’s Dispatches programme broadcast an episode called ‘Can AI Steal Your Vote?’

The premise was that 12 undecided households were told they were going to see some social media content that parties had been working on but had not released yet. First hook being they would be told something that others did not know, instant buy-in for most people.

What they were not told was that this was an experiment to see how people could be manipulated to vote in specific way based on information they were presented with. Any experimental social scientists might question the ethics of such an action, how could they know that long term harm could not be done to the subjects, but we will come to that later.

So, AI was used to have Rishi Sunak, and Sir Keir Starmer appear to make comments that they had not, examples being:

SUNAK: ‘To save the NHS we are going to introduce a £35 charge for GPS appointment.’

STARMER: ‘In order to work our way back to Europe I will introduce a ban on the singing of’ Land of Hope and Glory’ in Schools.’

The experiment portrayed different materials to different party voters. The result was shock and horror to the subjects. Other examples such as ‘Hot Mike’ moments, similar to the Gordon Brown ‘Bigoted woman incident’ from 2010.

AI was used to take as short as a 30-second clip of one of the leaders and using various techniques generate false images and videos of them saying things that were designed to cause confusion and alarm to their supporters in the experiment.

One AI generated fact about Angela Rayner was questioned by one of the subjects who then looked up the web to see if he could find any evidence that this was true, he found nothing.

At the end of the experiment subjects were asked to vote for Labour or Conservatives and then were told of the deception. All were shocked. To go back to the notion of harm, one female subject said it would be hard to believe anything she saw in the future.

There has already been evidence of Trump using AI to generate images that are fake. The BBC on the 4^th March this year exposed fake images used in his campaign of Donald Trump where he was pictured with a group of Afro-Caribbean voters. The BBC had reported back in February that the major High-Tech companies had agreed to tackle deceptive advertising using AI in political campaigns in the US.

The adage that America sneezes today and Europe catches a cold tomorrow comes to mind.

In the 2024 General Election campaign we have already seen instances where candidates have been dropped by their parties for comments, they claim they did not make. Could AI have played a part in these incidents?

With more and more use of social media being made for campaigning purposes the risk of interventions by fake AI generated materials is a growing risk. The ease of access to software and applications that can generate AI images and video is something we must be aware of.

You may have seen AI used in social media to take an image you load up and generate pictures of yourself as pirates, gangsters or other types of characters.

Comments and posts on social media are propagated around the internet exponentially. Your contacts see the posts, then their contacts and the contacts’ contacts. You can limit your posts to specific groups and friends but how much control do you really have. Candidates in this election have had posts for years ago thrown back at them. Nothing really goes away unless you delete it.

I would question the timing of Channel 4 broadcast before polling day. They did put watermarks over the false images and video, which included one of Stephen Fry advocating for both parties, when the source clip used did neither, but those images were powerful.

One of the subjects in the experiment said of the Fry clip they saw that this was a man of integrity and intelligence, so they believed what he was saying.

Rishi Sunak has long promoted the UK as a centre of excellence for AI, with great potential for good comes great potential for misuse.

What if AI is used to twist and generate false images, video and audio. One of the most shocking aspects of the programme, was one of the subjects was an expert in AI who was taken in.

* Les Tarr is a member of Banffshire and Buchan Coast Liberal Democrats.