AI-generated disinformation poses threat to UK general election integrity, CETaS report finds

A report published by the Alan Turing Institute’s Centre for Emerging Technology and Security (CETaS) warns that AI-generated disinformation could be used to undermine democracy ahead of the upcoming UK general election — and beyond.

The report, titled “AI-Enabled Influence Operations: The Threat to the UK General Election”, finds that while AI’s current impact on specific election results is limited, it poses broader risks to the democratic system.

These include a “degraded and polarised information space” and online harassment through deepfakes. AI could enhance these kinds of threats across the various stages of the UK general election cycle.

The report found that only 19 of 112 national elections since January 2023 showed AI interference. It also found no clear evidence that election results significantly differed from polling data.

However, the confusion created by AI-generated content has damaged trust in online sources, according to the report. Deepfakes have incited online hate against political figures, and politicians could exploit AI disinformation for electoral gain.

Earlier this year, London Mayor Sadiq Khan called for a crackdown on disinformation after a deepfake audio of his voice making inflammatory remarks before the UK’s Remembrance Weekend commemorations was leaked.

Though examples of AI misuse are scarce, they are often amplified through mainstream media, inflating public anxieties about AI’s threat to electoral processes.

The report is the first of two CETaS publications on AI and election security — the second will be published in September — and identifies three categories of election security threats.

Campaign threats that aim to manipulate voter behaviour or attitudes toward candidates or political issues; information threats that seek to undermine the quality of the information environment, confuse voters, and damage the integrity of electoral outcomes; and infrastructure threats target the systems and individuals responsible for securing election processes — using tactics such as ‘hack and leak‘ operations and AI-generated phishing emails against election officials.

The report stresses the urgency of addressing ambiguous electoral laws on AI use, which both domestic and foreign actors could exploit. For example, political parties might misuse AI to fabricate campaign endorsements, undermining the election process.

However, with the UK general election set for 4th July, there is limited time to enhance election security protections.

According to the report, “ambiguous electoral laws on AI use during elections are currently resulting in misuse.”

The Labour Party has expressed concerns about social media platform X’s refusal to remove deepfake audio clips of party leader Sir Keir Starmer from October 2023. Some of these clips have received 1.5 million views.

As such, CETaS calls for immediate actions, including setting more explicit expectations on AI use for political parties and media organisations and analysing data from recent UK local elections to inform contingency planning.

Further suggestions include issuing ‘fair AI use’ guidelines and voluntary agreements for political parties, supporting media with AI threat reporting tools, and launching public AI awareness campaigns.

CETaS has developed a timeline mapping potential AI threats to the UK general election and corresponding countermeasures.

The timeline shows when threats will likely emerge and what outcomes they aim to achieve from pre- to post-election. It’s based on evidence from recent elections and academic literature and estimates the time windows for interventions to mitigate these threats.

The report underscores a pressing need for a coordinated, “whole-of-government” approach to safeguard the integrity of the upcoming UK general election.

With the election date rapidly approaching, the window for implementing robust security measures is narrowing. CETaS concludes that immediate action is essential to mitigate AI-generated threats and ensure public trust in the democratic process.