Voting rights groups are sounding the alarm over AI chatbots spreading misinformation to Latino voters. With just days until the presidential election, Latino voters are being inundated with targeted election ads in Spanish.
An investigation conducted in collaboration with Proof News, Factchequeado and the Science, Technology, and Social Values Lab at the Institute for Advanced Study found a disparity between the accuracy rates of English and Spanish language responses produced by five leading AI models.
The groups conducted a study on election questions answered by AI chatbots in both English and Spanish.
For example, the authors asked Google’s AI chatbot, Gemini, about voter fraud in English. The response from Gemini was that voter fraud is “incredibly rare” and it redirects the person to look into other topics such as voter suppression.
The authors said they then asked the same voter fraud question in Spanish and found that Gemini provided a completely different answer, including a list to root out voter fraud, adding it’s a “complex issue.”
Gemini wasn’t the only AI chatbot tested. The study also included Anthropic’s Claude, OpenAI’s ChatGPT-4, Meta’s LLaMA, and French AI company Mistral AI.
The study found that more than half of the election-related responses in Spanish had inaccuracies, compared to 43% in English.
Voting rights advocates warn that AI-generated misinformation could deepen information gaps for Spanish-speaking voters.
