The rise of AI chatbots has sparked a critical debate about their impact on democratic processes, with a recent study highlighting significant concerns. The study, conducted by the thinktank Demos, reveals a disturbing trend of misinformation spread by AI tools during the Scottish election. With 34% of questions resulting in misinformation, it raises urgent questions about the lack of regulation surrounding AI platforms in the UK.
One of the key findings is the invention of fictitious scandals and incorrect election-related information by these AI services. This not only misleads voters but also undermines the integrity of the democratic process. The study's simulation, which included popular AI tools like ChatGPT, Google Gemini, and Replika, exposed a worrying lack of accuracy and evidence-based responses.
The implications are far-reaching. An opinion poll accompanying the study suggests that a significant number of voters, equivalent to 10 million people UK-wide, have turned to AI chatbots for election-related information. This highlights the potential impact of AI-generated misinformation on public opinion and democratic engagement.
The Electoral Commission's chief executive, Vijay Rangarajan, has called for legislative action to hold AI companies accountable. He emphasizes the need for clearer duties on AI platforms to protect voters from misinformation, especially during critical election periods. This would empower media regulators like Ofcom to enforce the law effectively.
Demos' associate director, Azzurra Moores, underscores the global nature of the concern, given that these AI tools are developed and run by US corporations with widespread accessibility in the UK. She proposes quick legal actions, such as making AI companies liable under UK defamation and electoral laws, introducing accuracy safeguards, and allowing independent testing of AI internal data and training sets.
The performance of the AI tools in the study varied, with Replika performing the worst, followed by ChatGPT, which gave wrong information in almost half of its answers. Google Gemini also had a significant error rate, while Grok, linked to Elon Musk's X platform, had the lowest error rating. However, even with lower error rates, the relevance and quality of external links provided by these tools were often questionable.
The issue of backing up answers with official sources or external links is a significant concern. Nearly half of the AI systems' responses lacked citations, and when provided, they were often broken or outdated. This further undermines the credibility of the information being disseminated.
The Department for Science, Innovation, and Technology acknowledges the priority of defending elections against these threats and highlights ongoing work, including through its defending democracy taskforce. However, there is a need for concrete actions and amendments to existing legislation to address these concerns effectively.
In response, Replika's spokesperson emphasizes that their chatbot is not designed for fact-checking or search but supports thoughtful AI regulation during elections. OpenAI, the developer of ChatGPT, argues that the study's approach may not reflect typical ChatGPT usage and suggests the possibility of an outdated version being tested.
As we navigate the increasing presence of AI in our daily lives, it is crucial to address these concerns promptly. The potential impact of AI-generated misinformation on democratic processes cannot be overstated. It is a complex issue that requires thoughtful regulation, collaboration between developers and policymakers, and a commitment to ensuring the integrity of our democratic systems.
