AI Chatbots Mislead U.S. Midterm Voters, Threatening Democratic Integrity

The upcoming United States midterm elections on November 8 will determine party control of both the House and the Senate. In recent years, voters have increasingly relied on artificial intelligence (AI) chatbots—such as ChatGPT, Bing AI, and Google Bard—to check voting schedules, candidate positions, and district information. These tools have attracted a massive user base with convenient features such as instant responses and voice interaction, and they are widely shared on social media and instant messaging platforms.

However, recent research shows that AI chatbots frequently make errors when providing election information, often confusing data from different sources and attributing facts to the wrong entities. Researchers discovered that some responses contained incorrect voting times, erroneous candidate profiles, and even treated disinformation generated by foreign actors as factual information. These errors not only diminish the credibility of the information, but they also risk forming false political perceptions among voters prior to voting, increasing the chaos surrounding election information.

Issues of this nature pose a dual challenge to the integrity of elections. First, misinformation undermines voters' trust in the voting process, which in turn affects voter turnout and voter behavior. Second, foreign actors utilizing AI to spread fake news could influence voter opinions undetected, further undermining democratic institutions. The U.S. Federal Election Commission and state-level election administration agencies have begun monitoring the impact of AI on election information and are discussing whether new regulatory frameworks for AI content are needed.

To address these challenges, AI companies and platform providers have begun strengthening content moderation. OpenAI, the developer of ChatGPT, has stated that it will improve its models' data sources and fact-checking capabilities, while cooperating with independent fact-checking organizations to provide "authenticity labels" to help users identify the credibility of information. Google and Microsoft have also stated that they will step up the screening of AI responses and add official source prompts for important election information. In addition, election administration agencies and civil society groups are collaborating to launch "AI literacy education" programs to teach voters how to distinguish between AI-generated content and genuine news.

In summary, while AI chatbots improve information accessibility, they also bring unprecedented cybersecurity challenges. To ensure the fairness and transparency of elections, the government, tech companies, and all sectors of society must work together to establish clear regulations and educational measures, ensuring that voters can make their voting decisions in an accurate and reliable information environment. (Source of facts: Central News Agency)