As though we haven’t got enough to deal with at the minute, social media has presented us with a new, endemic problem: AI-generated racists.
These online characters used to be easy to spot. They’d pop up on video feeds with three hands, talking out of their ears, as a swarm of blurred zombies wandered in the background.
But, in an alarmingly short period of time, they have become more realistic. The videos still show signs of inauthenticity – the racist avatars have artificial skin texture and the lighting is sterile – but the clues are a lot less obvious.
They are also dangerously cheap and easy to create.
The outcome is industrial-scale “AI slop”. Our social media feeds are swamped with fake videos purporting to depict interviews with “ordinary voters” complaining about rampant immigration, or preened news anchors spraying out false headlines about asylum seekers. People are even creating fake podcasts with AI-generated hosts and guests (Rory and Alastair beware).
Ergo, the cost of producing propaganda has dropped to near-zero.
If you wanted to create a far-right media channel before the advent of ChatGPT, you’d have to find someone dumb and prejudiced enough to go on camera, procure all the equipment, and then spend hours shooting, editing and promoting the content. This cost and effort acted as a major barrier – thankfully – to wannabe racist influencers.
Now, all you have to do is draft a few well-tailored AI prompts. Virtually anyone can start a far-right platform. Combine this with the in-built algorithmic incentives to pump out emotive content that appeals to people’s worst instincts, and we’ve somehow managed to turbocharge the worst aspects of an already dystopian social media climate.
Meanwhile, digitally-illiterate politicians are largely ignorant to the problem, their naivety encouraged by Big Tech lobbyists who want to block regulation.
I’m writing this article now, because I think the “deepfake” debate has been glaringly absent from the recent UK elections. Old school political commentators, who possess the same amount of digital savvy as their House of Commons counterparts are – unsurprisingly – missing this major evolution in our national news consumption.
For example, I recently saw an Instagram video where a chap on the street (a real chap, not an AI avatar) spewed out a bunch of random, false information about government freebies for asylum seekers.
Granted, he could have acquired these “facts” from the Daily Mail, The Sun or a myriad of other right-wing channels, all of which are renowned for spouting lies about brown people. But, I think it’s far more likely – given the outlandish nature of his grievances, and the fact no-one reads newspapers anymore – that he was fed these falsehoods on a Big Tech platform.
He may have seen some of the videos exposed by the Bureau of Investigative Journalism, which has been unmasking the people behind this deluge of racist AI slop. In one case, they unearthed a Sri Lankan influencer who’d made $300,000 running Facebook accounts pushing “racist, Islamophobic and anti-migrant” AI content to Brits. When questioned by the Bureau, the influencer said he targets “old people” because “they are the ones who don’t like immigrants”.
In another wild story, the Bureau revealed that a devout Muslim based in Pakistan is making a living from Islamophobic fake AI videos served to British audiences. One of his accounts even pushed out an AI-generated clip of Keir Starmer in Islamic dress, saying: “We are still a proud Muslim country. Our crazy violent Muslim immigrants, including our precious criminals and rape gangs, are the bedrock of our society.”
Grim.
Meanwhile, Full Fact tracked down a viral AI video, which had received 1.7 million views, showing a fake interview with two Muslim women saying people shouldn’t bring dogs to the beach. The video generated a “huge amount” of negative comments online, according to the fact checker.
This is how people now consume the news – through 10 second rage bait manufactured by AI and fine-tuned to trigger their prejudices. And, amid the fury and noise of social media – our hypnotising, insomnia-inducing TikTok feeds – people struggle to sift the real from the fake.
Receive all my latest essays and investigations…
I’m one of the first people to admit that the legacy media system is failing. Most big platforms are dependent either on a broken advertising-based business model or on billionaire backers. Neither is a recipe for good journalism.
However, the mass redundancies that have pillaged the British press over the past two decades have created a vacuum that has been filled with grifters. We’ve replaced trained journalists with content creators who wield AI to mass produce fake news.
The results are scary – and I’ve been confronted with them first-hand. The dad of one of my best friends used to be an avid newspaper reader – poring through The Times, Guardian and Financial Times every weekend. Now, he’s a Facebook addict in need of a digital detox – absorbing half-baked conspiracy theories and belching them out over the dinner table. “I need someone to save my husband from Facebook,” my friend’s mum said to me.
It’s difficult to comprehend – or put into context – the full scale of this problem, because social media feeds are walled gardens. There’s no central repository that allows us to see the content that’s going viral, and the people it’s reaching. While you can pick up a copy of The Sun – if you have the patience – and see the misinformation glaring back at you in red and black, social media content is elusive. Each of our feeds is completely different, which means we’re living in distinct – if occasionally overlapping – online realities.
However, even if we don’t know the depth of this social media-induced sepsis, the prognosis doesn’t look good.
From the little we know, racist AI slop is primarily being deployed by con artists to make money. Most social media platforms pay content creators for the amount of views they generate, and it turns out some of the biggest profits come from serving racist AI videos to British Facebook boomers.
More worryingly, there are reports of deepfakes being used in foreign influence operations – a trend that will surely only grow as AI platforms “improve”.
We recently published a DeSmog story about a Sri Lankan network pumping out pro-oil AI videos on UK Facebook feeds. In response to the piece, Phil Newell from Climate Action Against Disinformation said that AI slop was the “new Cambridge Analytica scandal”.
I agree with him.
That scandal involved shady consultants unlawfully profiling voters in order to target them with tailored political messages. Racist AI slop is similar in many ways. For the purpose of making money, and sometimes swaying votes, it uses underhand online tactics to push people towards radical politics.
However, rather than the carefully designed PR operation executed by Cambridge Analytica, AI slop is an onslaught – a barrage of hysteria overwhelming our collective immunity to falsehoods.
And, because there’s no single, centralised body behind the content, it’s almost impossible to suppress. The accounts can be shut down, but they immediately respawn with new names and carry on spewing out the same bile.
This problem seems intractable, but addressing it is fundamental to the survival of our democracy. If people continue consuming news via social media – which seems inevitable – and those feeds are saturated with racist hallucinations, our society will quickly rupture. The clamour will grow (as we are seeing) for far-right, fascist parties that feed those illusions – turning Britain into a place of intense racial and religious loathing.
In my mind at least, Britain is generally a tolerant place. The internet isn’t, and it’s steadily remaking our country in its image.
About me
I investigate the rich, populist and powerful – focused on Nigel Farage and his cronies. I’ve written for the likes of the New York Times, The Guardian, and New Statesman, have worked with Led By Donkeys, and have written two books on elitism and inequality.
Impatience: killer of governments
The most consequential outcomes of last week’s elections appear to be the fall of two administrations – the end of Welsh Labour’s 100-year rule, and Keir Starmer’s (significantly shorter) time in Dow…
For me, it's the feeling of hopelessness to do anything about the rate of dumbification of society. Millions of people consuming lies and nonsense at a huge volume and at breakneck speed. Or perhaps at breaksociety speed. Education, education, education.
Truly scary times.