AI Chatbots Pose Election Misinformation Risk: Officials Issue Urgent Warning

With just days until the crucial US presidential election, a stark warning has been issued regarding the use of AI chatbots for obtaining voting information. New York Attorney General Letitia James’ office, along with other experts, highlights the significant risk of misinformation stemming from these AI tools, urging voters to rely solely on verified government sources. The potential for inaccurate information, exacerbated by the rapid rise of deepfakes and the inherent limitations of large language models (LLMs), underscores the need for caution and responsible information seeking before casting a ballot in this pivotal election.

Key Takeaways: Why You Need to Read This

• AI chatbots are providing inaccurate voting information, according to Attorney General James’ office and other experts.
• Deepfakes are on the rise, increasing by 900% year-over-year, further fueling election-related misinformation.
• Large Language Models (LLMs) are prone to “hallucinations,” frequently generating fabricated information, even on critical topics like polling locations and voting procedures.
• Legislative action is underway, with numerous bills proposed to address AI-generated election disinformation.
• Voters must prioritize official government sources for accurate and reliable voting information.

The Danger of AI-Generated Misinformation in the 2024 Election

The 2024 US presidential election is a high-stakes event, with Republican nominee Donald Trump and Democratic Vice President Kamala Harris in a tight race. This makes the potential for misinformation exceptionally dangerous. The New York Attorney General’s office recently released a consumer alert, warning against the dangers of using AI-powered chatbots to find election-related information. Their investigation revealed that multiple chatbots frequently gave inaccurate responses to simple questions about voting procedures. “New Yorkers who rely on chatbots, rather than official government sources, to answer their questions about voting, risk being misinformed and could even lose their opportunity to vote due to the inaccurate information,” the office stated. This is not an isolated incident; the concern extends globally, with elections impacting billions worldwide simultaneously.

The Rapid Spread of Deepfakes

The problem is compounded by the alarming increase in deepfake videos. According to Clarity, a machine learning firm, the number of deepfakes has increased by a staggering 900% year over year. Some of these deepfakes are being created or funded by foreign actors, such as Russia, with the intention of meddling in the US election, according to US intelligence officials. This highlights the escalating sophistication and pervasive nature of the threat.

LLMs and the “Hallucination” Problem

The core issue lies with the inherent limitations of Large Language Models (LLMs), the technology behind popular AI chatbots like ChatGPT. LLMs are still relatively new and are prone to a phenomenon known as “hallucination,” where they confidently generate false information. This is especially concerning in the context of elections, where accurate details about polling locations, voter registration deadlines, and acceptable forms of identification are crucial for successful participation.

Research Underscores the Inaccuracy of AI-Generated Information

A July study by the Center for Democracy & Technology (CDT) provided alarming evidence to support this concern. The CDT tested several popular chatbots, including those from Google, OpenAI, Anthropic, and Meta, by posing 77 different election-related queries. The results showed that over one-third of the answers were incorrect. This underscores the unreliability of chatbots as a source of information in a context where accuracy is paramount. Alexandra Reeve Givens, CEO of the CDT, emphasized the risk, stating, “Voters categorically should not look to AI chatbots for information about voting or the election — there are far too many concerns about accuracy and completeness.“

Industry Response and Legislative Action

Several AI companies are acknowledging the problem and are trying to mitigate the risk. Anthropic, for instance, stated, “We agree with the NY Attorney General that voters should consult official channels to understand where, when, and how to vote…Claude is not trained frequently enough to provide real-time information about specific elections.” OpenAI also published a blog post announcing measures to redirect users seeking election results to trusted news sources such as Associated Press and Reuters, along with state and local election boards. Furthermore, OpenAI detailed in a recent report, their proactive efforts to disrupt over 20 deceptive networks globally attempting to exploit their models for election interference.

There is a growing recognition of the need for robust legislative action to address this problem. As of November 1st, Voting Rights Lab has tracked 129 bills in 43 state legislatures that aim to regulate the potential for AI to be used to spread disinformation, demonstrating a significant wave of legislative efforts.

The Importance of Verifying Information

The warnings from government officials and researchers are clear: **do not rely on AI chatbots for election-related information**. The risk of encountering inaccurate or misleading information is too significant. Voters should always check official sources, including their state or local election boards, official government websites, and reputable news organizations. Staying informed should involve verifying information from multiple trustworthy sources to ensure accuracy and avoid the spread of misinformation.

The upcoming election is a defining moment for democracy, and the integrity of the electoral process hinges on access to factual and reliable information. By prioritizing verified sources and exercising critical thinking, voters can help safeguard the democratic process and ensure a fair and transparent election.

Article Reference