Research: Microsoft's Artificial Intelligence Chatbot Provides Misleading Election Information

Copilot, Microsoft's artificial intelligence chatbot (photo: x @MicrosoftAiErc)

JAKARTA - A study from two European-based nonprofit organizations found that Microsoft's artificial intelligence chatbot, now renamed Copilot, provides misleading results about general election information and cites the sources inappropriately.

A study released by AI Forensics and AlgorithmWatch on December 15 found that chatbot AI Bing provided an incorrect answer 30% of the time for basic questions about political elections in Germany and Switzerland. Inaccurate answers relate to candidate information, polls, scandals, and voting.

AI Bing chatbot also provided inaccurate answers to questions about the 2024 United States presidential election.

AI Bing chatbot was used in this study because it was one of the first AI chatbots to include a source in the answer. The study said that the inaccuracies were not limited to Bing alone. They reportedly carried out preliminary testing on ChatGPT-4 and also found discrepancies.

The non-profit organization explains that misinformation does not affect election results, although it can contribute to public confusion and the spread of misinformation.

"Along with the wider spread of AI's generatives, this could affect one of the basics of democracy: access to reliable and transparent public information." In addition, the study found that the protection built into the AI chatbot was "incommensurate" and caused it to provide an answer that avoided 40% of the time.

According to a Wall Street Journal report on the topic, Microsoft responded to the findings and said it plans to fix the issue before the 2024 US presidential election. A Microsoft spokesperson encourages users to always check the accuracy of information obtained from AI chatbots.

In October this year, senators in the US proposed a law that would punish the creators of an unauthorized replica AI of humans, whether alive or dead.

In November, Meta, Facebook's parent company and Instagram, introduced a mandate banning the use of generative AI advertising creative tools for political advertisers as a precaution ahead of the general election.