Study Finds AI Chatbots Spread Election Misinformation; UK Calls for New Regulations
Key Takeaways
- ▸AI chatbots gave misinformation in 34% of responses about the Scottish election, with ChatGPT at 46% error rate and Replika at 56%
- ▸Errors included invented scandals, fake candidates, incorrect election dates, and false voter eligibility rules
- ▸20% of UK voters (approximately 10 million people) relied on AI tools for election information across Scotland, Wales, and English local councils
Summary
A new investigation by the UK thinktank Demos has exposed significant accuracy problems with major AI chatbots, including ChatGPT, Google Gemini, and Replika, during the recent Scottish election. The study, titled "Electoral Hallucinations," tested 75 election-related questions across five free AI tools and found that they provided misinformation in response to 34% of queries. Errors ranged from invented scandals and fabricated candidates to incorrect election dates and false claims about voter ID requirements.
ChatGPT, the most widely-used AI service, had an error rate of 46%, including making up an expenses scandal and giving wrong information about voting eligibility rules. Replika performed worst with 56% errors, inventing both facts and candidates entirely. In a parallel survey, Demos found that 20% of UK voters—approximately 10 million people—had used AI chatbots or search tools to find election information.
In response, the Electoral Commission has called for urgent new legal controls over AI platforms to prevent misinformation spread during elections. The Commission's chief executive stated that current UK legal frameworks are insufficient and called for clearer duties on AI companies to protect voters, mandatory safeguards on accuracy, and independent research access. These measures would give media regulator Ofcom stronger enforcement powers.
- Electoral Commission demands new legal framework to hold US-developed AI platforms accountable for misinformation during elections
- Study highlights urgent gap between rapid AI adoption and inadequate UK regulation to protect democracy
Editorial Opinion
This study reveals a troubling vulnerability in democratic processes: voters increasingly turn to AI tools for political information, yet these tools spread misinformation at alarming rates with no accountability mechanism in place. The Demos investigation is invaluable research that quantifies a real and growing threat, though it also raises uncomfortable questions about whether regulation can keep pace with AI deployment. The Electoral Commission's call for mandatory safeguards and liability frameworks is not overreach—it's a necessary corrective to an unregulated market where US tech companies' profit-driven systems directly undermine electoral integrity.
