An AI voice clone that holds live conversations: Raman campaign accuses pro-Bass PAC of impersonation

By KTLA · October 26, 2026.

The campaign of Nithya Raman, a candidate for mayor of Los Angeles, filed complaints on Monday with California's Fair Political Practices Commission (FPPC) and the city's Ethics Commission. Both target Thrive LA PAC, a political action committee that supports Mayor Karen Bass. A PAC is an organization that raises and spends money to influence elections.

Raman's campaign says Thrive LA has put up billboards around the city urging people to call a phone number presented as Raman's. According to her team, callers reach an AI-generated voice clone that speaks in the first person as the candidate and holds a live conversation, making up answers to voters' questions. The complaint says the call does not disclose that it is a machine and does not identify who is paying for the service.

Raman's campaign alleges violations of several California laws. One is Section 20012 of the Elections Code, which prohibits maliciously distributing digitally fabricated audio that makes a candidate appear to say something they did not say within the 30 days before an election. Another is Section 84514 of the Government Code, which requires committee ads with content generated or substantially altered using AI to include the statement “Ad generated or substantially altered using artificial intelligence.” The campaign also cites alleged failures to meet disclosure requirements.

Raman spokesperson Liz Garcia called the conduct “sad and illegal.” According to the campaign, the line is funded by “private equity executives and conservative donors of Thrive LA.” The Bass campaign had not responded to the allegations when KTLA published its story. All of the above are allegations from one side: the available material includes no ruling by any agency and no response from Thrive LA. KTLA adds that some social media users were outraged by the billboards, while others treated them more as a joke.

Our take: what matters is not only the case itself but the technical category. The laws cited in the complaint refer to “fabricated audio” and “ads,” concepts designed for a recorded piece that gets distributed. A clone that generates new answers on every call is something else: a conversational agent speaking with a real person's voice. According to Raman's campaign, that difference is the crux, because the AI “does not merely play a fabricated recording” but generates new statements. It remains to be seen whether regulators apply those laws to this format or whether there is a gap.

More broadly, the technology to clone a voice from a few minutes of audio and converse with low latency is within anyone's reach, and a candidate with hundreds of hours of public speeches is ideal material. That shifts the risk: the problem is no longer the fake viral video, which can be debunked, but thousands of private, one-on-one conversations that nobody audits. The most effective defense is traceability: every AI interaction should disclose that it is AI and say who is paying for it.

In the short term, this points to campaigns with more impersonation, more complaints and more distrust of any voice on the phone. It is an uncomfortable transition, and the rules will need to be tightened and updated based on evidence, without regulating out of panic. In the long term, the same technology now used to deceive will let any citizen consult a verified assistant about each candidate's actual proposals. Getting there requires one basic rule: AI can speak, but it has to say who it is and who sent it.