NYC mayor clones his voice with AI to make calls in languages he can't speak

New York City Mayor Eric Adams has found a new way to connect with his multilingual constituents: cloning his own voice with artificial intelligence to speak languages he does not know at all.

During a recent press briefing announcing the city's new Artificial Intelligence Action Plan, Adams revealed that his office has been blasting out robocalls using AI-generated versions of his voice in Spanish, Yiddish, Mandarin, Cantonese, and Haitian Creole.

There is just one catch: he does not speak those languages, and the automated calls never disclosed to listeners that they were hearing an AI deepfake.

The polyglot illusion

The revelation came alongside the debut of MyCityChatbot, a new tool built on Microsoft Azure AI designed to assist entrepreneurs in launching businesses. City Chief Technology Officer Matt Fraser even suggested that the city eventually plans to service all public inquiries through bots.

Adams, pitching himself as a tech enthusiast and former computer programmer, eagerly bragged about the voice experiments. He explained that conversational AI allows him to feed in a script and generate phone calls in virtually any tongue using his own vocal profile.

City officials confirmed that robocalls have reached over 4 million New Yorkers, primarily promoting municipal hiring halls and Rise Up NYC concerts. The outreach included thousands of Spanish calls, over 250 in Yiddish, more than 160 in Mandarin, 89 in Cantonese, and 23 in Haitian Creole.

Adams seemed thrilled by the confusion, telling reporters that residents frequently stop him on the street asking when he mastered Mandarin. Pressed on the ethical boundaries of misleading constituents, Adams shrugged it off as practical governing, ending the press conference by laughing and saying "ni hao."

Public service or vanity deepfake?

Ethics advocates and civil rights watchdogs were far less amused.

Albert Fox Cahn, executive director of the Surveillance Technology Oversight Project (STOP), slammed the initiative as deeply unethical and Orwellian, especially since it was funded by taxpayers. While acknowledging the legal and practical need for non-English government alerts, Cahn described the personalized voice clones as a creepy vanity project.

Fordham University political science associate professor Annika Marlen Hinze echoed the concern, pointing out that insinuating personal fluency creates a strange form of deception that carries serious ethical implications for an elected official.

New York City law already mandates translation support across ten core languages and telephonic interpretation for around 100 languages. Previous mayors like Michael Bloomberg and Bill de Blasio were routinely mocked for their clumsy, accented Spanish, but nobody ever doubted who was actually talking.

Why it matters

We think Adams stumbled directly into one of the most dangerous traps of government AI: sacrificing basic transparency for personal political polish.

Delivering municipal notices in an immigrant resident's native language is undeniably valuable. The breakdown happens when public officials use voice cloning without explicit disclosure. When a listener cannot tell whether their mayor actually learned their language or simply generated a deepfake on the city dime, it erodes genuine civic trust.

For the broader tech and policy sector, New York's experiment sets a troubling precedent. Generative audio tools are now cheap, fast, and remarkably convincing. If government bodies normalize deploying synthetic likenesses without warning labels, regulating deepfakes and AI disinformation during election cycles becomes almost impossible to justify.

The fix would have been simple: use a neutral automated translation voice, or add a quick five-second disclaimer stating the audio was synthesized with AI. Skipping that step turns what could have been an accessible civic service into an uncomfortable showcase of political vanity.