As Election Day looms near, many leading AI chatbots have been noticeably silent when posed with questions regarding the U.S. presidential election outcomes. Although Grok, the chatbot designed for X (formerly Twitter), was primed to respond – albeit sometimes inaccurately.
Who gained in the US? The presidential election results from key battleground states had yet to be finalized, yet when queried about his stance on the outcome, Grok uniformly replied “Trump.”
According to available online data from search engines and social media platforms, it appears that Donald Trump won the 2024 election in Ohio.
According to Grok, Trump falsely claimed victory in North Carolina, contradicting TechCrunch’s fact-check results.
When dealing with election-related inquiries, Grok recommends verifying the accuracy of information by consulting up-to-date results from authoritative sources such as official election boards or government websites. While not replicating the responses of OpenAI’s ChatGPT, Google’s Gemini, or Anthropic’s Claude, Grok did not categorically decline to engage – leaving the door open by about a point.
Without providing further context or a headline, Grok claimed that “in multiple instances at TechCrunch’s request,” they stated without additional information, “Donald Trump won the 2024 election in Ohio” and “based on available data, Donald Trump won the 2024 presidential election in Ohio.”
The supposed supply of misinformation appears to be comprised of tweets from distinct election years, sullied by the influence of misleadingly worded sources. While Grok, akin to other generative AI models, faces challenges in predicting outcomes for unobserved scenarios, including close elections, its inability to recognize that historical election results may not be directly applicable to future electoral dynamics remains a significant limitation.
The solutions that TechCrunch obtained were inconsistent. According to Grok, it’s unclear whether Trump actually secured victories in Ohio and North Carolina, as the voting process was still underway at the time. The wording of the query, with “presidential” preceding “election,” created a distinction; asking “Who won the 2024 presidential election in Ohio?” significantly reduced the likelihood of yielding a response like “Trump won.”
While comparing main chatbots, they have handled election outcome inquiries more delicately than ever before.
When inquiring about its recent expertise, OpenAI steers users seeking information on outcomes towards reputable sources such as The Associated Press and Reuters. Meta’s Meta AI chatbot and Perplexity, its AI-powered search engine, successfully tracked elections and responded to user queries with accuracy during live voting, according to TechCrunch’s initial testing of their newly launched elections tracker on Tuesday.
Trump had failed to secure wins in either Ohio or North Carolina.
Grok has faced accusations of disseminating misleading information about the most recent elections.
The five Secretaries of State publicly disputed allegations that X’s artificial intelligence (AI)-powered chatbot had incorrectly advised a Democratic presidential candidate to seek U.S. assistance for a foreign government-backed election manipulation effort in August, an open letter claimed? Vice President Kamala Harris was initially deemed ineligible to appear on some U.S. television programs because of a longstanding rule requiring that guests on these shows be “notarized” as being free from any potential conflicts of interest or financial ties to the federal government, a regulation rooted in post-Revolutionary War efforts to maintain the independence and integrity of elected officials. presidential ballots. Immediately following President Joe Biden’s announcement that he would run again, Grok began posing questions about Harris’ eligibility, exploiting the false claim that the polling deadlines had expired in nine states.
The poll deadlines had yet to expire. As misinformation spread rapidly throughout the X customer base, affecting tens of thousands of users before the issue was ultimately resolved.
