Deepfakes, disinformation and elections

Three billion people go to the polls this year across the world, including in the US and the UK. Bot creators are having a field day. In the US, there are reports of Russian-based fake online accounts spreading disinformation about President Biden and the Democrats in order to influence the outcome of the election. A fake ‘robocall’ purporting to be from President Biden was uncovered in New Hampshire, telling people not to vote in the primaries. A fake AI recording of Sadiq Khan circulated late last year, saying he didn’t give a ‘flying shit about the Remembrance weekend,’ and ‘What’s happening in Gaza is much bigger.’ An AI-generated recording of Kier Starmer swearing at his team had 1.5 million hits on X in 24 hours.

And in January, German digital forensic experts uncovered a massive pro-Russia disinformation campaign using around 50,000 fake accounts on X to send around two messages per second over a one-month period, designed to undermine public support for Ukraine. The tone of the AI-generated messages was similar to that of the AFD, the far-right party in Germany.

Communicators face a massive challenge in cutting through this noise, and asking people to question the validity of information they read online.

AI scams on the rise

There’s an amazing story on CNN about a finance worker for an unnamed 'multinational firm’ who paid $25 million of company funds to fraudsters after attending a video conference call with what she thought was the company’s UK CFO and a number of her colleagues. In fact, they were all deepfakes, and she was the only ‘real’ person on the call. She made a number of transactions as requested by the CFO, and only realised it was a scam after checking in with the company HQ.

These kinds of scams will only increase, according to the NCSC, which says AI will make it harder to identify what’s real and what’s a scam. AI makes phishing messages more convincing by mimicking the tone and style of real messages from colleagues, so eliminating the errors that used to make it easier to spot a scam.

Reputation starts with treating people decently

X has finally paid severance pay to the employees it sacked in Ghana more than a year ago after the acquisition by Elon Musk. During the layoffs at X, Musk said all employees got three months’ pay. Not so for the Ghana team - some of whom had moved with their families to Ghana to work for X – who had to take legal action to reach a settlement agreement. The agency representing them, Agency Seven Seven, said X only agreed to discuss their pay after the BBC covered the story.

We’ve seen this with the Post Office – it seemingly takes national or international media coverage to shame an organisation into fulfilling even the most basic legal obligations to its people. If you care about your reputation, start by caring about your people.