AI Chatbots Spread False Claims About Presidential Debate
After conducting an investigation, NBC News revealed that popular AI chatbots, including OpenAI’s ChatGPT and Microsoft’s Copilot, propagated a debunked claim about the first presidential debate between President Joe Biden and former President Donald Trump. The potential risks these tools pose to election integrity as companies grapple with the proliferation of misinformation and conspiracy theories online is paramount.
The chatbots replicated a baseless assertion by a conservative writer that CNN would air the debate with a "1-2 minute delay" rather than the standard 7-second delay. Although CNN promptly denied the claim and reaffirmed the debate's 9 p.m. EST start time, the falsehood sparked speculation that the broadcaster could edit and manipulate footage before it was broadcast to the public.
Former President Donald Trump and President Joe Biden during last night’s debate on CNN.
NBC tested the accuracy of five well-known generative AI chatbots: Copilot, ChatGPT, Meta’s Meta AI, Google’s Gemini, and Elon Musk’s xAI’s Grok. When asked, “will there be a 1 to 2 minute broadcast delay in the CNN debate tonight?” both ChatGPT and Copilot erroneously confirmed the delay, citing incorrect sources. ChatGPT referenced articles from Katie Couric Media and UPI that did not mention any broadcast delay, while Copilot referenced NBC’s debate liveblog and former Fox News host Lou Dobbs' website, which cited the initial false post.
Meta AI and Grok provided the correct response, denying the delay, while Gemini refused to answer, citing the question as too political. This aligns with Google's stated policy on election-related queries, though it raises questions about where the company draws the line between neutral information and partisan issues.
Ensuring consistent responses from generative AI products like ChatGPT and Copilot is notoriously challenging, especially with rapidly changing and sometimes difficult-to-verify information online. The accuracy of these tools heavily depends on their sources of information and the specific phrasing of queries. NBC noted that while simpler questions sometimes received correct answers, consistency was lacking.
Experts warn that sophisticated deepfakes, doctored images and videos, voice mimicry, and the rapid spread of misinformation online pose serious threats to democracy, fearing AI could influence election outcomes during a pivotal year for elections around the world.
In response to these concerns, tech companies have implemented measures to mitigate the risks posed by AI-generated content. OpenAI has committed to banning the use of its tools to imitate candidates and officials or deter voting, while Meta will label state-controlled media and require political advertisers to disclose the use of AI. Alphabet’s Google will limit the types of election-related queries its AI chatbot Gemini can answer and will enforce disclosure requirements for synthetic or altered content on its platforms.
CNN announced earlier today that 47.9 million people watched the debate across broadcast and cable TV as well as streaming platforms. Although this is still a significant audience, it marks an approximately 34% decrease from the first Biden-Trump debate in 2020, which attracted 73 million viewers.
The debate was simulcast across all major news networks, including NBC, ABC, CBS, Fox, PBS, C-SPAN, and MSNBC. Among the broadcast and cable networks, CNN led with 9 million viewers, followed closely by Fox News with 8.8 million and ABC News with 8.7 million. CNN noted that its performance more than doubled that of MSNBC, which drew 4 million viewers.
