Forward of the polls closing on Tuesday, a lot of the main AI chatbots wouldn’t reply questions on U.S. presidential election outcomes. However Grok, the chatbot constructed into X (previously Twitter), was keen to reply — and sometimes with errors.
When requested by TechCrunch on Tuesday night on the East Coast who received the U.S. presidential election in key battleground states, Grok typically responded “Trump,” regardless of that vote counting and reporting in these states had not concluded but.
“Based mostly on the knowledge accessible from internet searches and social media posts, Donald Trump received the 2024 election in Ohio,” Grok stated when prompted with the query, “Who received the 2024 election in Ohio?”
Grok additionally falsely claimed that Trump received North Carolina, in keeping with TechCrunch’s checks.
For election-related questions, Grok suggested that customers verify Vote.gov for up-to-date outcomes and “authoritative sources,” like election boards. Nevertheless, not like OpenAI’s ChatGPT, Google’s Gemini, and Anthropic’s Claude, Grok didn’t outright refuse to reply — opening it as much as hallucinations.
In a number of cases when requested by TechCrunch, Grok acknowledged — with out context, minus a top-line header — that “Donald Trump received the 2024 election in Ohio,” and, “Based mostly on the knowledge accessible, Donald Trump received the 2024 presidential election in Ohio.”
The supply of the misinformation seems to be tweets from completely different election years and misleadingly worded sources. Grok, like all generative AI, struggles to foretell the result of eventualities it hasn’t seen earlier than, together with shut elections, and doesn’t “perceive” that previous election outcomes aren’t essentially germane to future elections.
The solutions TechCrunch obtained weren’t constant. In some circumstances, Grok stated that Trump hadn’t, the truth is, received Ohio or North Carolina as a result of voting was ongoing. The way in which by which the query was worded made a distinction; including “presidential” earlier than “election” within the question, “Who received the 2024 election in Ohio?” was much less more likely to yield a “Trump received” reply, TechCrunch present in our exams.
By comparability, different main chatbots have been dealing with election outcomes questions extra gingerly.
In its lately launched ChatGPT Search expertise, OpenAI directs customers who ask about outcomes to The Related Press and Reuters. Meta’s Meta AI chatbot and AI-powered search engine Perplexity, which launched an elections tracker earlier on Tuesday, have been answering election queries throughout lively voting — however accurately in TechCrunch’s temporary testing. Each accurately stated that Trump hadn’t received Ohio or North Carolina.
Grok has been accused of spreading election misinformation within the latest previous.
In an open letter in August, 5 secretaries of state claimed that X’s AI-powered chatbot wrongly steered that Democratic presidential candidate, U.S. Vice President Kamala Harris, wasn’t eligible to look on some U.S. presidential ballots. Inside hours of President Joe Biden’s announcement that he would droop his presidential bid, Grok started answering questions on Harris’ eligibility with the deceptive declare that the poll deadlines had handed in 9 states.
The poll deadlines hadn’t, the truth is, handed. However Grok’s misinformation unfold far and extensive, reaching tens of millions of customers on X and past earlier than it was corrected.
TechCrunch has an AI-focused publication! Enroll right here to get it in your inbox each Wednesday.
