In one of the zanier twists of this campaign season, two Virginia congressional candidates will face off tonight, in a virtual debate, against an AI bot.
The candidates are Bentley Hensel and David Kennedy, both running for Virginia's 8th district -- both independents, and both expected to lose.
The bot is based on Rep. Don Beyer, the Democratic incumbent, heavily favored to win, who decided not to show up when the True Representation Movement and the Black Alumni National Alliance invited him to the only debate in his district.
Hensel, one of the candidates, is also the software engineer who built the bot. He told DFD he developed it without Beyer's permission out of frustration that the lawmaker appeared to be sitting out campaign events. "Sadly, [it's] not that unusual for an incumbent congressperson in a safe seat," he said in an interview.
The debate has already gotten some coverage for the AI angle, though the whole event is a bit of a stunt -- the True Representation Movement is run by Ahmed Bouzid, a Virginia activist and businessperson who officially joined Hensel's team as campaign manager last week. Bouzid sees the debate itself as an effort to "highlight the fact that incumbents take their seats for granted" and to "showcase how AI can [be] used responsibly."
Beyer's office declined to comment. Republican candidate Jerry Torres is also likely to sit out, according to hosts of the debate. He did not respond to a request for comment.
Tonight, the plan is for moderators to have access to the chatbot and type in their questions. The bot will provide answers as if Beyer were answering. (There's an irony to the bot being built around Beyer, a 74-year-old Virginia pol who has gotten public attention for actually going back to school to learn more about AI technology.)
Hensel said he used generative AI to train a bot on Beyer's press releases, official websites and Federal Election Commission data. He couldn't find enough audio to generate a convincing Beyer voice, so Hensel said a generic AI voice will provide responses.
The specter of an unauthorized AI bot being built to dunk on a candidate's absence raises an obvious question: Is this really OK?
Using generative AI to develop someone's likeness isn't a new issue, and it's not illegal. AI replicas of famous people, including some notable psychologists and academics, already exist -- and have raised some alarms about the lack of guardrails around this issue.
Although some lawmakers have introduced bills to regulate AI in elections, Congress appears unlikely to pass any of them ahead of Election Day. And even that legislation was intended to regulate deepfakes and other technologies intended to spread misinformation -- which the Beyer bot is not. Without comprehensive policies in place, nothing is stopping Hensel, or others like him, from making more chatbots like this one.
Political bots have been built for parody or entertainment value before. Chatbot2024, a separate entity from any candidate, used AI to generate chatbots of the Republican presidential candidates last year. The Twitch channel TrumporBiden2024 hosted a long-running AI-generated nonsense debate between Donald Trump and Joe Biden.
Nathan Sanders, a data scientist and affiliate at Harvard University's Berkman Klein Center for Internet and Society, said he has seen candidates deploy AI largely to represent themselves, and that candidates using the technology to develop chatbots of their opponents raised new ethical questions.
"I think disclosure is important here. It seems very clear in this case that they are disclosing that it's an AI. And it's not actually Beyer on the stage," he said.
Hensel said he trained the AI to be as "unbiased as possible," and to answer any questions accurately as Beyer. The Don Bot was originally developed on ChatGPT -- a product of OpenAI -- but Hensel said his account was removed from the site recently for violating its guidelines on political use. Now the Don Bot's coding is hosted on GitHub and powered by Cloudflare AI.
If Beyer were to commit to the debate, Hensel said, the Don Bot wouldn't be needed. "I mean, I'd be a little sad, after all that work," he said.
Hensel doesn't seriously think the debate will hurt Beyer or propel his own candidacy to new heights. "I'm not naive about what's gonna happen in November," Hensel said. "I'm well aware I'm not gonna win."
He does say he hopes his stunt could provide a template for pressuring other public figures to engage with constituents, the media and the people whose lives they affect.
Hensel's next project? POTUSGPT. Hensel said he is developing chatbots of Vice President Kamala Harris and former President Donald Trump ahead of Election Day. He said he'll train them on Harris and Trump's specific policy platforms and their records as politicians. With extensive examples of their voices in the public record, the bots should sound pretty convincing.
"I expect to be sued," he said with a laugh. "But how many people will never be able to talk with the president?"
The rhetoric around Elon Musk's stewardship of X is starting to get downright biblical.
Speaking with POLITICO's Sam Clark and Barbara Moens, outgoing European Commission Vice President Věra Jourová said Musk "is not able to recognize good and evil" and is, in fact, a "promoter of evil."
Jourová called X on Wednesday "the main hub for spreading antisemitism," saying "we are in the situation where the member states' law enforcement powers have to protect the people who are under threat, under physical threat... [from] this new chapter, new intensity of antisemitism, where we don't see sufficient action from the side of the platforms."