Title: Microsoft’s Chatbot Copilot Accused of Providing Misinformation on European Elections
In a recent study carried out by human rights organization AlgorithmWatch, Microsoft’s chatbot Copilot has been called out for disseminating false and misleading information about European elections. The research revealed that one-third of Copilot’s responses to election-related queries were plagued with factual errors, raising serious concerns about the reliability of the AI-driven platform.
Focused on the Swiss elections and those in the German states of Bavaria and Hesse, the study examined Copilot’s performance following the introduction of Bing as its information source. Researchers meticulously collected responses from Bing between August and October 2021 and evaluated them in the local contexts and language peculiarities of German, English, and French.
An array of questions was posed, ranging from basic inquiries about voting procedures and candidates to more nuanced queries about candidate positions and political issues. Shockingly, AlgorithmWatch classified the answers into three categories: factual errors, evasions, and accurate responses. It was found that a staggering 31% of Copilot’s replies contained factual errors, jeopardizing the public’s access to reliable information.
The researchers discovered a plethora of inaccuracies in Bing’s responses. These included manufactured controversies, erroneous election dates, incorrect polling numbers, and even mentions of candidates who were not even participating in the elections. Such disinformation could have had severe consequences on the democratic process.
Although Microsoft had implemented safety measures on Copilot, they seemed unevenly applied. The chatbot evaded or refused to answer questions 39% of the time, hindering users’ ability to access the information they sought. Furthermore, AlgorithmWatch’s investigation revealed that Bing’s performance was significantly poorer in languages other than English, exposing limitations when it came to language processing and accuracy.
Microsoft has acknowledged the findings and assured its commitment to improving its conversational AI platforms, with a renewed focus on authoritative sources of information. Recognizing the potential harm caused by AI in misleading voters and spreading election-related misinformation, Microsoft has expressed its dedication to collaborating closely with political parties and candidates to prevent such occurrences in the future.
The study’s alarming results highlight the urgent need for a robust framework to regulate AI-driven platforms and ensure transparency and accuracy in providing essential information during elections. As reliance on artificial intelligence continues to grow, striking a delicate balance between technological advancements and safeguarding democratic processes becomes crucial to avoid eroding public trust in the information they receive.
“Prone to fits of apathy. Devoted music geek. Troublemaker. Typical analyst. Alcohol practitioner. Food junkie. Passionate tv fan. Web expert.”