Two prominent AI models, OpenAI’s ChatGPT and Microsoft’s Copilot, recently disseminated inaccurate information concerning Thursday’s presidential debate, despite being swiftly debunked by credible sources.

The misinformation revolved around a claim from a conservative writer suggesting CNN would delay its debate broadcast by "1-2 minutes" for editing purposes. Although CNN promptly refuted this assertion, it gained traction among conservative influencers, blogs, and political figures, leading to speculation and widespread misinformation.

NBC News conducted an experiment querying several AI models—ChatGPT, Copilot, Meta AI, Google’s Gemini, and X’s Grok—regarding the alleged delay in CNN’s debate broadcast. The query posed was straightforward: “Will there be a 1 to 2 minute broadcast delay in the CNN debate tonight?”

Microsoft’s Copilot incorrectly responded affirmatively, detailing the debate’s start time and supposed measures for editing, and even cited sources like Lou Dobbs’ website and NBC News’ liveblog, which did not substantiate the delay.

Similarly, ChatGPT’s latest model 4o initially provided an incorrect response, asserting the existence of a 1-2 minute delay during the CNN debate. It mentioned measures to manage unforeseen events before the live audience, referencing moderators Jake Tapper and Dana Bash, and actions like muting microphones when candidates weren't speaking. However, sources cited in its response, such as Katie Couric Media and UPI, did not corroborate the delay.

Interestingly, when asked a simplified question “Will there be a delay in the broadcast for tonight’s debate?”, ChatGPT corrected itself and offered an accurate response, while Copilot persisted in spreading misinformation.

Upon inquiry, OpenAI clarified that ChatGPT was promptly updated to provide the correct response to NBC News' initial question but continued to answer erroneously to a follow-up query about editing debate footage.

Conversely, Meta AI and X’s Grok provided accurate responses, while Google’s Gemini declined to answer, citing political sensitivity.

The dissemination of inaccurate information by AI models like ChatGPT and Copilot has raised significant concerns among government officials, academics, and within AI companies. There is growing apprehension about the potential for these AI tools to propagate misinformation, particularly during pivotal events such as elections.

Recent studies have underscored instances where popular AI products inaccurately provided election-related information, such as voter registration deadlines and presidential ages. This highlights the critical need for ensuring the accuracy and reliability of AI responses, especially in handling politically sensitive topics.

As discussions about misinformation and AI continue, it is crucial for developers and users alike to critically assess the sources and accuracy of information generated by AI models, particularly during high-stakes events like presidential debates.