A robocall that imitated President Joe Biden and told some New Hampshire voters not to vote in the state’s primary election has become a sharp example of the risks around AI voice cloning. The person or group behind the call has not been identified, but two separate teams of audio experts told WIRED the recording was likely created with technology from ElevenLabs.
The finding matters because ElevenLabs is not a fringe tool. It is a prominent AI voice startup whose technology is marketed for uses such as audiobooks and video games, while also being available through a paid service that lets users clone a voice from an audio sample.
What researchers found in the Biden robocall
Pindrop, a security company that builds tools to identify synthetic audio, analyzed a 39-second clip of one of the calls. In a blog post, the company said the audio pointed to ElevenLabs’ technology or a “system using similar components.”
The company compared patterns in the clip against more than 120 different voice synthesis engines. Pindrop CEO Vijay Balasubramaniyan said the team did not expect a clear match because tracing the origin of AI-generated audio can be difficult. Instead, he said, “It came back well north of 99 percent that it was ElevenLabs.”
Pindrop also checked its approach against audio samples known to have been created with ElevenLabs’ technology, and against another voice synthesis tool. The goal was to test whether the method could distinguish the suspect call from other generated audio.
ElevenLabs itself offers an AI speech detector on its website that says whether an audio clip was made with the company’s technology. When Pindrop tested the robocall sample with that detector, the result came back as 84 percent likely to have been generated using ElevenLabs tools. WIRED said it independently received the same result using Pindrop’s sample.
A second team reached a similar conclusion
Hany Farid, a digital forensics specialist at the UC Berkeley School of Information, was not immediately convinced that ElevenLabs was the source. His first reaction was based on the sound quality of the call, which he felt did not match what he expected from the company’s technology.
Farid said, “When you hear the audio from a cloned voice from ElevenLabs, it’s really good.” He also said, “The version of the Biden call that I heard was not particularly good, but the cadence was really funky. It just didn’t sound of the quality that I would have expected from ElevenLabs.”
But after his team at Berkeley conducted its own independent analysis of the audio sample obtained by Pindrop, it arrived at the same broad view. Farid said, “Our model says with high confidence that it is AI-generated and likely to be ElevenLabs.”
Taken together, the Pindrop and Berkeley analyses do not identify who made the call. They do, however, point toward a likely technology source and show that specialist teams may be able to investigate the origin of synthetic political audio after it spreads.
Why ElevenLabs is central to the debate
ElevenLabs has grown into one of the best-known companies in AI voice cloning. It recently raised $80 million at a $1.1 billion valuation in a funding round co-led by Andreessen Horowitz. According to the company’s CEO in a recent Bloomberg article, investors value ElevenLabs at more than $1.1 billion.
Its investors include Andreessen Horowitz, Sequoia Capital, SV Angel, Nat Friedman, former CEO of GitHub, and Mustafa Suleyman, cofounder of AI lab DeepMind, now part of Alphabet. That level of backing has made the company one of the most visible players in synthetic voice technology.
The company’s safety policy says it is best to get permission before cloning someone’s voice. The same policy also says permissionless cloning can be acceptable for several non-commercial purposes, including “political speech contributing to public debates.”
In a statement provided late Friday, ElevenLabs CEO Mati Staniszewski said the company is “dedicated to preventing the misuse of audio AI tools,” but cannot comment on specific incidents. The statement also said ElevenLabs acts on misuse cases, including by assisting authorities.
The Biden robocall is not the only political misuse concern connected to the company’s tools. Last September, NewsGuard, which tracks online misinformation, claimed that TikTok accounts spreading conspiracy theories with AI-generated voices, including a clone of Barack Obama’s voice, used ElevenLabs’ technology.
At the time, ElevenLabs said in an emailed statement to The New York Times, “Over 99 percent of users on our platform are creating interesting, innovative, useful content,” while also saying it recognized misuse and was developing safeguards to limit it.
The larger risk for elections and public trust
The central problem is not only that AI-generated voices can sound convincing. It is that the tools are broadly available, while reliable verification is still difficult for people outside specialist fields.
ElevenLabs and similar voice tools can support legitimate uses, including creating audiobooks more cheaply. But the same availability also creates openings for malicious political content, impersonation, and fast-moving misinformation.
Sam Gregory, program director at the nonprofit Witness, which helps people use technology to promote human rights, described the challenge plainly. “We have a real problem,” he said. “When you have these very broadly available tools, it's quite hard to police.”
A Discord server for ElevenLabs enthusiasts has included discussion of plans to clone Biden’s voice, along with links to videos and social media posts featuring deepfaked Biden content or AI-generated copies of Donald Trump and Barack Obama’s voices.
For election officials, journalists, and voters, the timing problem may be just as serious as the detection problem. A careful forensic analysis may eventually point toward an AI system, but that process can take longer than the time it takes for a deceptive recording to influence people.
Gregory warned that the necessary tools are not yet broadly available to the people who may need them most. “Journalists and election officials and others don't have access to reliable tools to be doing this quickly and rapidly when potentially election-altering audio gets leaked or shared,” he said. “If this had been something that was relevant on election day, that would be too late.”
What the robocall case makes clear
The Biden robocall shows both sides of the AI audio problem. Researchers may be able to analyze a suspicious clip and identify signs of a particular voice synthesis system. But that expertise is not easy to access quickly, and ordinary listeners cannot reliably determine where a recording came from.
If the Pindrop and Berkeley analyses are correct, the call was made using technology from one of the most prominent and well-funded AI voice startups. That makes the question of safeguards more urgent, especially as synthetic audio becomes easier for companies and individuals to experiment with.
The case is not only about one call, one company, or one election message. It is a warning about the gap between what AI voice cloning can now produce and what institutions can quickly verify when deceptive audio enters public debate.