Report Reveals AI Chatbots Sharing False Information on EU Election

According to a study, chatbots developed by Google, Microsoft, and OpenAI are suspected of spreading incorrect information concerning the European election. This material was reportedly provided just two months before people were scheduled to cast their votes.

According to a non-governmental organization (NGO) headquartered in Berlin that conducted the study in March, artificial intelligence tools tended to produce erroneous election schedules and details about how to cast a vote. This trend occurred despite the programs maintaining their political neutrality. According to the findings, chatbots often supplied links to YouTube movies or information in Japanese that were either broken or irrelevant.

Researchers observed that artificial intelligence chatbots were dynamic, making duplicating the experiment difficult, but the experiment was sufficiently big to function as a representative sample. Ahead of the June 6th to June 9th European election, it also revealed fresh information regarding the hazards of so-called artificial intelligence hallucinations, which often emerge due to limited training data, biases, and erroneous assumptions.

During the course of the test, the most successful version of ChatGPT was the most recent paid version, while Google’s Gemini was regarded as having the lowest probability of providing accurate responses.

There has been a substantial amount of anger and criticism directed against Google’s Gemini AI picture generator, which has been accused of displaying biases and producing images that are erroneous or insulting.

One of the most prominent issues was that Gemini produced photographs that depicted historical characters, such as the Founding Fathers, popes, and Nazis, as being of different races, even though this was not the case in reality. Even when users requested historically accurate portrayals, Gemini appeared to resort to a dream-world strategy of visuals.

Google’s Gemini was reportedly in the midst of implementing corrections back in March, as stated by the company officials. According to what was said, the firm will continue to take prompt action when Gemini is not behaving in an acceptable manner.

According to Robin Koch, a representative for Microsoft, the company is working to fix difficulties, and preparations are being made to ensure that their technologies will function up to expectations for the elections in 2024. As an additional precaution, he said that some of the efforts included providing users of Microsoft’s Copilot with election information derived from reliable sources and encouraging them to carefully examine online connections.