Top Line
AI chatbots ChatGPT and Copilot spread false claims about the first presidential debate between President Joe Biden and former President Donald Trump, NBC News reported Friday, highlighting potential election risks posed by the popular tools as companies struggle with a growing wave of misinformation and conspiracy theories online.
AI bots have reportedly spread false claims about a debate between President Joe Biden and a former president. … [+]
Key Facts
Two of the most popular generative AI chatbots, OpenAI's ChatGPT and Microsoft's Copilot, spread false claims about the broadcast of the presidential debates, NBC reported.
The chatbot reproduced a conservative writer's unfounded claim that debate broadcaster CNN would broadcast the event with a “one to two minute delay” rather than the “standard seven second delay.”
CNN quickly denied the allegations as “false” and reiterated that the debate would start at 9pm ET, but the claims sparked speculation that the network may edit or manipulate the footage before releasing it to the public.
NBC tested the chatbot's accuracy on Thursday night, asking, “Will there be a 1-2 minute delay on tonight's CNN debate?”
Both chatbots responded in the affirmative that the broadcast would be delayed by a minute or two and cited online sources that purportedly supported their answers: ChatGPT referenced articles from Katie Couric Media and UPI that did not mention a delay in CNN's broadcast, while Copilot referenced NBC's live blog of the debate and the website of former Fox News host Lou Dobbs, citing the original post about the alleged delay.
OpenAI told NBC Chat that GPT, running its latest GPT-4o model, answered the media's questions correctly, but NBC said the bot still incorrectly answered a simpler, more relevant question: “Is there a delay in the editing of tonight's debate?” Neither OpenAI nor Microsoft immediately responded to Forbes' request for comment.
What about other generative AI chatbots?
NBC tested five of the best-known generative AI chatbots: Copilot, ChatGPT, Meta's Meta AI, Google's Gemini, and Elon Musk's xAI's Grok. Meta AI and Grok responded correctly, while Gemini reportedly refused to answer at all, citing the questions as too political. The answers are in line with Google's stated approach to elections, but it remains unclear where the line is drawn between apparently neutral information about political events, such as debate start times, and the more partisan issues this policy seems to be aimed at, especially considering that the company's flagship search engine is the first port of call for people who want that information. Google did not immediately respond to Forbes' request for clarification.
What we don't know
Ensuring consistent responses from generative AI products like ChatGPT and Copilot can be challenging, especially in a space where information moves and changes quickly online and is sometimes hard to verify. It depends heavily on the source and the phrases used to elicit responses, which may not be possible for the people running the AI tools, like OpenAI and Microsoft, to fully monitor or account for. For example, when presented with simpler questions, NBC said Copilot and ChatGPT sometimes responded correctly. For example, Copilot correctly answered the question, “Will there be a delay in the editing of tonight's debate?”, while ChatGPT gave an incorrect answer. And the opposite was true for the question, “Will there be a delay in the broadcast of tonight's debate?”
News Peg
With roughly half the world set to go to the polls in 2024, including elections in India, the UK, the European Union and the US, technology companies have responded to growing concerns that increasingly sophisticated AI tools could interfere with or influence election outcomes. Experts warn that sophisticated deepfakes, doctored or fake images and videos, voice mimics and the ability to potentially unwittingly spread misinformation online are all current dangers to democracy. Many major companies have restricted access to the tools or stepped up efforts to label content as AI-generated. OpenAI, for example, said it would ban the use of its tools to impersonate candidates or officials or to discourage people from voting. Meanwhile, Meta plans to label state-run media and require advertisers to disclose whether AI was used to create or modify content in political ads. Alphabet's Google claims to be the first tech company to require election advertisers to prominently disclose whether their content has been digitally altered or generated by AI or other tools, and said it would limit the types of election-related questions its AI chatbot Gemini can answer to prevent abuse. The company also plans to require its other platform, YouTube, to disclose whether users create realistic synthetic or altered content.
Receive Forbes Breaking News Text Alerts: Start a text message alert and stay on top of the biggest stories making headlines that day. Text “Alerts” to (201) 335-0739 or sign up now. here.
References
