Electoral Commission Seeks New Rules on AI Chatbot Misinformation

The Electoral Commission called for new legal controls on AI chatbots after a Demos study found the tools provided inaccurate information in 34 percent of responses about the recent Scottish election. Demos tested five free AI services with 75 questions across three constituencies. The tools invented scandals, placed candidates in the wrong races, and gave the wrong election date.

Replika produced errors in 56 percent of answers, including a fabricated expenses scandal and an invented candidate. ChatGPT gave incorrect information in 46 percent of responses, including an inaccurate election date and voter eligibility rules. Google Gemini was wrong in 22 percent of cases, including a claim that a police fraud investigation into the Scottish National Party was ongoing.

Grok recorded the lowest error rate at 9 percent, though its links were often irrelevant.

An opinion poll of 2,005 British adults found that 20 percent had used AI chatbots or search tools for election information, equivalent to 10 million people nationwide. Vijay Rangarajan, the Electoral Commission’s chief executive, said voters want accurate information and that the current legal framework should go further.

He called for clearer duties on AI platforms to protect voters from misinformation. U.S. corporations raises concerns without a legislative framework to protect the public. The Department for Science, Innovation and Technology said defending elections against these threats is an absolute priority and that work is ongoing across government.