evidoria

← Back to browse

Good practice Imported

UK Electoral Commission's AI Deepfake Detection Pilot — Human-Reviewed Monitoring Ahead of the May 2026 Elections

United Kingdom · London · See the United Kingdom profile · See the London profile

Evidence: Descriptive / self-reported Top 74% 47/100 · Ask Evidence Copilot about this practice

Ahead of the 7 May 2026 elections in England, Scotland and Wales, the Electoral Commission ran an AI-assisted pilot to spot deepfake audio and video about candidates, with every flagged match checked by a human analyst before any takedown request.

~25%
UK adults who recalled seeing a 2024-election deepfake
30%
UK adults who saw an election deepfake (post-2026-election poll) (Apr-May 2026)
16%
UK adults who encountered a deepfake 5+ times (Apr-May 2026)

Details

Maturity
Pilot
Promoter
The Electoral Commission (UK), with the Home Office
Period
Apr–Jun 2026
Keywords
elections, democratic integrity, AI-assisted monitoring, public communications

Context

Following its own polling on the 2024 UK general election — where over half of voters recalled seeing misleading information about parties or candidates and around a quarter reported seeing or hearing a deepfake — the Electoral Commission sought a tool to spot synthetic election-related content ahead of the 7 May 2026 elections in England, Scotland and Wales.

Activities

Run from April to June 2026 in partnership with the Home Office, the AI-assisted pilot monitored online content for synthetic audio and video intended to mislead voters about the electoral process or falsely depict candidates; every match identified by the system was reviewed by a human analyst before the Commission worked with parties, candidates or platforms to correct or remove material.

Results

Independent post-election polling (30 April-6 May 2026) found 30% of UK adults said they had seen a deepfake or AI-generated image, audio clip or video about a candidate or politician, and 16% said they had encountered one more than five times.

Conclusions

The Commission said it would publish pilot findings after the May elections, with results not expected for at least six months, so this remains an early, unevaluated deployment.

Implementation
Indicative cost
Medium (€50k–€500k)
Time to results
Short (< 1 year)
Staffing & skills
Electoral Commission analysts (human review of every AI-flagged match), Home Office partnership

Conditions for success

  • Every automated match reviewed by a human analyst before any correction/takedown action, limiting false positives from imperfect detection technology

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Similar practices you may find useful