Shortly after Joe Biden announced he was ending his re-election bid, misinformation began to spread online about whether a new candidate could take the president’s place.
Screenshots claiming that a new candidate could not be included on ballots in nine states quickly spread across Twitter — now X — and racked up millions of views. The Minnesota secretary of state’s office began receiving fact-checking requests for these posts, which were completely wrong: The deadlines for ballots had not passed, giving Kamala Harris plenty of time to have her name added to the ballots.
The source of the misinformation: Twitter’s chatbot Grok. When users asked the AI tool whether a new candidate still had time to be included on the ballot, Grok gave the wrong answer.
Finding the source and working to fix it served as a test case for how election officials and AI companies will interact during the 2024 U.S. presidential election, amid fears that AI could mislead or distract voters. And it showed the role that Grok, in particular, could play in elections — as a chatbot with fewer barriers to prevent the generation of more incendiary content.
A group of secretaries of state and their representative organization, the National Association of Secretaries of State, contacted Grok and X to report the misinformation. But the company did not work to correct it immediately, instead giving the equivalent “I think it’s fair to say that we all thought that was the wrong answer,” said Minnesota Secretary of State Steve Simon.
Fortunately, this misguided response was not too risky: it would not have prevented people from voting, but the clerks quickly took a tough stance on what might happen next.
“We thought, what if the next time Grok makes a mistake, it’s about something bigger?” Simon said. “What if the next time the answer is wrong, he’s asking if I can vote, where I vote, what are the hours, or if I can vote absentee? That alarmed us.”
Particularly worrying was the fact that the social media platform itself was spreading false information, rather than users spreading misinformation using the platform.
The secretaries made their initiative public. Five of the group’s nine secretaries signed a public letter to the platform and its owner, Elon Musk. The letter called on X to have its chatbot adopt a similar stance to other chatbot tools, such as ChatGPT, and direct users who ask Grok election-related questions to a trusted, nonpartisan election information site, CanIVote.org.
The effort worked. Grok now directs users to a different website, vote.gov, when asked about elections.
“We look forward to keeping the lines of communication open during this election season and stand ready to address any additional concerns you may have,” Wifredo Fernandez, X’s head of global government affairs, wrote to the secretaries, according to a copy of the letter obtained by The Guardian.
It was a victory for the clerks and for curbing election misinformation, and a lesson in how to respond when AI-based tools aren’t enough. Calling out misinformation early and often can help amplify the message, give it more credibility and force a response, Simon said.
Although he was “deeply disappointed” by the company’s initial response, Simon said: “I want to say congratulations and thanks as well, and that’s what they deserve. This is a great company, with a global reach, and they decided to do the right and responsible thing, and I congratulate them for that. I just hope they keep it up. We’ll keep following up.”
Musk has described Grok as an “anti-woke” chatbot that gives “spicy” responses often laden with sarcasm. Musk is “against centralized control to any degree that it’s possible,” said Lucas Hansen, co-founder of CivAI, a nonprofit that warns about the dangers of AI. This philosophical belief puts Grok at a disadvantage in preventing misinformation, as does another feature of the tool: Grok pulls in the most prominent tweets to inform its responses, which can affect its accuracy, Hansen said.
Grok requires a paid subscription, but has the potential for widespread use since it’s built into a social media platform, Hansen said. And while it can give incorrect responses in chat, the images it creates can also further inflame partisan divisions.
Images can be extravagant: a Nazi Mickey Mouse, Trump flying a plane into the World Trade Center, Harris in a communist uniform. A study by the Center for Digital Hate Counting claims Grok can create “compelling” images that could fool people, citing images it prompted the robot to create of Harris taking drugs and Trump sick in bed, according to The Independent. reportedAl Jazeera media outlet wrote in a recent investigation that was able to create “realistic images” of Harris with a knife in a grocery store and Trump “shaking hands with white nationalists on the White House lawn.”
“Now any random person can create something that is substantially more provocative than they could do before,” Hansen said.