“The voice of my new and old AI,” she called it in a recent video introducing it to voters.
Wexton made headlines earlier this year when she spoke on the House floor using a robotic voice app, which was widely praised as a show of mental toughness. It wasn’t a voice like hers.
As Rep. Wexton stood up to address the House Appropriations Committee this week, she demonstrated her new, more natural-sounding voice, occasionally tapping away at the walker she has been using since being diagnosed with a Parkinson’s-like condition called progressive supranuclear palsy (PSP).
“If you’ve heard me speak before the PSP took over my voice, you might think your ears are being tricked now,” she told the committee through an app playing on her iPad. “Not so. Today we’re using a new AI model of my voice, and it’s absolutely amazing.”
The cadence, tone and timbre were all very similar to Rep. Wexton, who served five years as a Virginia senator before being elected to the House of Representatives in 2018. She is not seeking reelection this fall. It’s because of her health.
For her staff, the moment was more than just cool.
“That was a big deal, something I never thought I’d hear again,” Wexton public relations director Justin McCartney said Friday.
McCartney said Wexton, 56, has retained all of his wit and intelligence, but a neurological disorder makes it difficult for him to communicate verbally. “This voice has given back something to many people that they didn’t realize how much they had lost,” he said.
“It’s by no means ‘me,’ but it’s more of me than I or anyone around me ever thought we’d ever hear again,” Wexton said in written responses to questions from The Washington Post on Friday. She said she had been forced to decline speaking engagements and public appearances because of a lack of confidence in her voice, but that the technology has restored her ability to “continue to do my best in this job I love.”
The first thing she heard was a snippet of Hamlet’s “To be or not to be” monologue. “My husband broke into a huge smile,” she says. “I had never seen him be so generous and so serious before. [happy] “For so long,” she said, adding that she enjoys seeing the reactions of her friends and colleagues.
“It was so moving to hear my friend’s voice again,” Rep. Abigail Spanberger (D-Va.) said in an email Friday. “I know this new voice is computer-enabled, but I know the wise words and witty, yet biting jokes that come from it are Jennifer’s alone.”
Late last year, when Wexton was preparing to speak on the House floor in support of a bill to eradicate Parkinson’s disease, she wrote a speech and had Rep. Jennifer McClellan (D-Va.) read it for her. It was a humbling moment, McClellan told The Washington Post in an interview on Friday.
“For an independent person who’s built her career on using her voice for others, first as a prosecutor and then as a legislator, it’s been incredibly tough to have to rely on someone else to speak on her behalf,” McClellan said. “I think the AI software has given her some power back in a really moving way.”
Wexton acknowledged the pitfalls of AI in comments to The Washington Post, saying it’s “scary to think that someone with bad intentions could use this technology to do bad things.” He limits who on his team has access to the tool, because “if they use my voice to say something without my consent, that could really cause problems.”
Experts are increasingly sounding the alarm about the technology’s potential to disrupt politics: With national elections scheduled for this year around the world, a January report from the Wilson Center warned that “the risk of ‘blurring the walls of reality,’ as one analyst put it, through the use of AI-generated or more traditional deepfake creations is disturbingly high.”
The Brennan Center for Justice at New York University Law School noted late last year that generative AI is already being used in U.S. political ads and mailings and is “poised to redefine the modern campaign,” but the center’s report noted that Congress has yet to address the issue.
McClellan said Wexton’s implementation of AI highlights the need to update laws and regulations.
Wexton’s voice model was created after a video of her using a text-to-speech app caught the eye of New York-based voice AI startup Eleven Labs in May. The company reached out to the congresswoman’s staff, who gave the go-ahead. Over the course of a few weeks, the staff put together more than an hour of audio clips of Wexton from before her illness left her speechless.
It took the company just a few days to generate a digital version of Wexton’s voice, and an ElevenLabs spokesman said the artificial intelligence not only allows it to mimic a voice, but also to adjust the tone of the voice to sound natural and not robotic.
“Our models are able to understand the relationships between words and adjust pronunciation based on context to produce realistic, human-like speech,” ElevenLabs spokesman Sam Scholar said in an email.
McCartney said Wexton’s firm pays ElevenLabs a small subscription fee for the service.
ElevenLabs, which was founded in 2022, has made similar offers to other notable people, including former Spirit Airlines CEO Ben Baldanza. Lori Cohen, a litigation lawyer at Greenberg Traurig who has amyotrophic lateral sclerosis, is arguing cases in court using an AI version of her voice.
“Clearly there is work to be done to adequately protect against the potential dangers that it poses,” Wexton said in the video, “but it can also provide Americans with disabilities with new and life-changing opportunities they may never have imagined.”
