The Pulse
UK Watchdog Targets Three-Hour Alerts for Election Deepfakes
Electoral Commission chief executive Vijay Rangarajan said the watchdog is developing a system intended to alert candidates and parties to significant deepfakes within three hours. A Commission review of its May 2026 pilot reports separate
AI.info Team · Written with AI ·
“If that happens to the British election, that will move a significant amount of votes,”
Vijay Rangarajan, chief executive of the Electoral Commission, said.
Rangarajan was referring to a deepfake circulated during Ireland’s 2025 presidential election, which falsely showed candidate Catherine Connolly saying the vote had been called off and that she was withdrawing. At a panel at the Conservative Party conference, he said the UK’s Electoral Commission was developing an AI-assisted system intended to alert candidates and parties to significant deepfakes within three hours of their appearance on social media. Tech Policy Press reported his remarks on October 15, 2026.
Three hours to reach candidates
In the account of Rangarajan’s remarks, the alert window is intended to give candidates, parties and other authorities time to respond before a fake spreads widely. He said three hours is the kind of time it takes content to go viral. The plan he described is for AI to scan posts in real time and flag potential cases to Commission staff, who would review them and decide whether they meet the threshold for notifying affected candidates or parties.
The system is not designed to settle disputes between candidates over campaign claims. Rangarajan described the Commission’s approach as “perpetrator neutral”: instead of making the slower task of identifying and prosecuting the creator its first priority, the regulator would focus on alerting those affected. The three-hour period is a target described in his remarks, not a guarantee that every deepfake will be found or assessed within that time.
Three million posts, two sets of pilot figures
The Commission tested the approach during the May 2026 local elections. At the conference, Rangarajan said the AI processed about three million posts, flagging 147 items for human review; staff judged ten “really quite misleading” and referred them to social media companies or the police. He said the cases had not had a large effect, while arguing that the pilot gave the Commission a system it could use to help voters, media and political parties.
A Commission review published on September 14, 2026, describes the same pilot with different counts: about three million posts were processed, and human review identified 219 political deepfakes from 60 accounts. Those posts drew 9.43 million views and 403,000 likes, shares or comments; 15 met the Commission’s threshold for risk to electoral integrity. The review does not explain how those figures map to Rangarajan’s tally of 147 items and ten referrals, so the two sets should not be treated as equivalent measures.
The Commission’s review found no evidence of deliberate use of highly convincing political deepfakes to target or interfere with the May elections. It also stresses that an AI-generated image, audio clip or video is not harmful simply because it was made with AI: context, timing and audience matter. Staff must judge the content’s potential electoral effect, and the report says that human-led assessment takes specialist knowledge and limits how much material can be reviewed.
Platforms still control removals
The pilot does not give the Electoral Commission power to remove posts itself. Where material raises serious concerns, the regulator can request removal from a platform or report suspected legal breaches to police, but the Commission’s review describes uneven responses from social media companies. It says manipulated content reported to X on polling day, falsely alleging threats to candidate safety at polling booths, remained online despite appearing to breach the platform’s own rules.
That gap shapes what the three-hour alert described by Rangarajan can achieve: it can speed up notice and referrals, but it cannot compel a platform to act. The Commission says political deepfakes often fall into a regulatory grey area unless they meet specific legal conditions, and recommends updating electoral law and online-platform duties. For now, its response relies on human review and requests to platforms that retain control over whether flagged content comes down.