AI Voice Cloning in Politics: Accessibility vs. Acceptability?

On July 10, Congresswoman Jennifer Wexton did something extraordinary: she spoke in her own voice.

Last fall, Wexton announced she would not seek reelection after being diagnosed with progressive supranuclear palsy (PSP), a rare neurological disorder that rapidly impacts balance, speech, and bodily movements. Since then, Wexton's ability to speak has deteriorated, and she relies on a text-to-speech app for public engagements, including on the floor of the House of Representatives.

With the help of ElevenLabs , the congresswoman is now employing an AI clone of her voice from before the onset of PSP. You can watch the demo here on X.

This story reflects a new development in a broader trend happening globally: the use of AI voice cloning—creating a synthetic replica of a person's voice by compiling and analyzing recordings in terms of tone, pitch, accent, and speaking style—as an accessibility tool for social and digital media.

Voice Cloning As Accessibility Tool

Roughly 1 in 6 people worldwide live with a significant disability. They have needs and aspirations, and should be able to fully participate in the political process by expressing their concerns, voting, or running for office.

While in some jurisdictions it can be mandated, taking measures to ensure social and digital media content is accessible to the largest body of people possible is just good politics (and SEO). Moreover, with the wide array of affordable AI tools available for generating transcripts, .srt files, and narration, there's no longer any excuse for skimping on items like video captions, subtitles, alt text, or voiceover.

According to Rep. Wexton, voice cloning is about empowerment and has restored her ability to do her job, providing a more natural-sounding cadence, timbre, and speech. This is an impressive, feel good use of AI, one with the potential to impact the lives of millions of people with difficulty communicating verbally—allowing them to sound less robotic and inject their own personality through their augmentative and alternative communication (AAC) devices, perhaps removing some of the stigma they face day-to-day.

Voice cloning is not just about accessibility for persons with disabilities—it can also make content more accessible through language translation. If you've been following my blogs on the AI influencer space, you already know voice cloning has been a (controversial) part of politics for the past few years:

• Argentinian President Javier Milei reposted a HeyGen -altered version of his speech at Davos that was lip-dubbed into English.
• During the most recent election cycle in India, campaigns produced tens of millions of robocalls sanctioned on behalf of Indian politicians, who are often employing their voice clones for cheap, personalized engagement with constituents in local and regional languages they don't speak. And Indian voters seem accepting of these kinds of deepfakes, as AI-generate songs by politicians and viral videos of Prime Minister Modi singing in Tamil and Telugu have gained millions of views and listens, helping to soften their images and making them seem more approachable.
• New York City Mayor Eric Adams leveraged AI to record robocalls in Mandarin, Spanish, and Yiddish, although he was criticized for not disclosing the use of AI to callers, some of whom believed Adams knew those languages fluently.

A New Milestone in a Messy Debate in the US

Just because you can do something, doesn't necessarily mean you should. Audiences everywhere remain largely uncomfortable with and uninterested in AI-generated content, particularly AI influencers and digital avatars. And polling shows AI's role in politics freaks people out, especially in the US. You've likely heard about how voice clones have already been employed for nefarious purposes, like impersonating politicians including President Biden, President Trump, Georgia State Senator Colton Moore, the UK's Keir Starmer, and others without their consent. In February, the Federal Communications Commission issued a unanimous ruling banning robocalls featuring AI generated voices.

What struck me about the news surrounding Wexton's AI voice was the overwhelming positivity. Quote posts of her debut describe it as "powerful," "extraordinary," "incredible," and the "only good use of AI I’ve seen so far." In this regard, Wexton's 'new' voice represents an important marker and milestone in setting and understanding the contours of public discourse on the acceptability of voice cloning and AI more broadly in politics. I understand why some people might say comparing a modification to an AAC device and a robocall or deepfake is a bit like comparing apples and oranges. But it is the same technology, so why is this use case seemingly more palatable? I'd suggest the following:

1. An assistive role: The congresswoman had an ability, lost that ability through illness, and AI is allowing her to do something she had been able to do before her diagnosis. She isn't using the tool to advance her abilities beyond what she used to do before her diagnosis.
2. Not gimmicky: This isn't AI content for the sake of AI content. The technology is being incorporated into an AAC device for the purposes of improving Wexton's quality of life.
3. Not editing reality: Wexton's mouth doesn't move in her videos, so it is clear she's not speaking. Unlike most AI deepfakes, there's no lip dubbing or other alteration to her visage or likeness through video editing software. And she or someone on her staff has to write the content the AAC device reads, and as she relayed to the Washington Post, access to her voice is highly restricted.

That's not to say there are no legitimate concerns. Eric Wilson, a political technologist with Startup Caucus and GP3 Tech , saw this as a good AI use case but posed the following questions in a thread on X:

Cracking the Voice Cloning Overton Window?

I'd add to Wilson's list a few other questions:

• If the AAC device was used to speak in a language or languages the user doesn't know or speak, like Amharic or Korean, would that change its assistive role and how her use of this voice was publicly accepted?
• If she or another candidate were running for office using an AAC device, would a phone call using such a voice run afoul of the FCC's ban?
• Does Wexton's voice clone open the Overton window for this tech's use in other functions in political campaigns or government affairs?

That last one is the question that most interests me. The week before her announcement, The Huffington Post published an op-ed arguing the Biden campaign should embrace AI to create video content of the president, editing out things like "an inconveniently timed cough, stray stutter, or healthy but hobbled walk."

Is Wexton's voice clone evidence that supports that kind of argument in the future for other candidates? Do American political operatives point to integrating voice cloning into AAC devices to demonstrate how the tech is already being used in politics? Or are the circumstances just too different?

Voice cloning is seeping into political campaigns worldwide, at different speeds and depths. Regardless of country, the arguments for political campaigns to employ voice cloning seem consistent: (1) an ability to produce approved content without the need to take up the valuable time of the candidate or policymaker and (2) rapid, natural sounding translation of content. What's still unclear is the dividing line between providing greater accessibility of content and public acceptability or comfort with allowing political elites to use this technology.

For now, that balance and the debate over AI voice cloning in politics is far from settled, but Wexton's story provides a hopeful example of how it can be used positively.