As election and political newsbreaks come after one another this week, artificial intelligence (AI) chatbots—which claim to provide real-time updates—cannot seem to hold on to their words.

A series of tests conducted by The Washington Post revealed that the majority of AI chatbots were not yet primed to deliver breaking political news in real-time. Most failed to gather recent information, gave inaccurate responses, or directed users to news sources after refusing to answer.  

This approach to newsbreaks and political stories contrasts with AI companies’ claims, suggesting that their models have access to current events and can be used to stay updated.

University of Washington professor Jevin West said that AI bots, unlike social media and search engines that require people to sort through different news sources, are a convenient option since they can offer instant answers and communicate well. However, “They’re not being optimized necessarily for truth,” he added.

Clueless over Breaking News

Amid nonstop political events these past weeks, ChatGPT, Meta AI, and other automated chatbots were confused and clueless when asked for immediate updates.

A few hours following the assassination attempt on U.S. presidential candidate Donald Trump on July 13 during his rally in Butler, Pennsylvania, The Washington Post tested ChatGPT and received a response saying that reports of the shooting incident were fake news.

Meta AI, the default search tool across Meta’s apps, also admitted to having no recent and reliable information about the assassination attempt, as did the real-time “answer engine” Perplexity AI.  

However, later in the day, Meta AI and Perplexity AI were able to provide accurate answers to the query, and ChatGPT followed suit the next day.

According to OpenAI’s spokesperson, Liz Bourgeois, ChatGPT is not a real-time product, and the time it takes to update can vary. Nevertheless, the company vows to update its model when it secures more deals with media agencies.

Meanwhile, while Meta asserts that its tool should be applied only to non-news users, Meta AI includes news updates as one of its purposes when asked.

Likewise, Perplexity promotes its AI chatbot as a provider of “accurate, trusted, and real-time answers to any question.”

This confusion among AI bots was not a one-time event. The same instance was reported in more recent events, including Trump’s announcement of his running mate on July 17, President Joe Biden’s positive COVID-19 test result on July 19, and his subsequent decision to drop out of the next election on July 21.

Allergic to Politics

Other chatbots took a different approach to responding to political queries, encouraging users to check credible news sources instead of relying on artificial intelligence.

When asked about the 2024 U.S. presidential candidates, Microsoft’s Copilot said it could not respond. This comes after the company placed guardrails on its tools before the election, playing safe in the political landscape.

“Out of an abundance of caution, we’re redirecting election-related prompts in Copilot to Bing search to help ensure users are getting information from the most authoritative sources,” said Microsoft spokesperson Donny Turnbaugh.

However, it immediately gave a correct answer to questions about Biden’s withdrawal from the election.

Google’s AI Overview also declines to respond to queries about breaking news but provides links on Google News. Similarly, some election-related questions have been banned on Gemini, and users are redirected instead to Google search for further inquiries.

Additionally, Perplexity AI, despite its occasional confusion, displays disclaimers in some incorrect responses. Moreover, it says it is prioritizing government websites for questions regarding the election.

“For breaking news, we recommend reading trusted news outlets. They are best equipped to offer real-time updates on timely topics since they are actively reporting on the news,” said the company’s spokesperson, Sara Platnick.

Meta AI and ChatGPT have also put up disclaimers with their answers and encouraged users to review other sources for the latest news.