The telephone has always been a place where the human voice matters. You hear a tone, a hesitation, a familiar cadence, and your guard comes down. In the weeks before the November midterm elections, that vulnerability has become a political problem. Anna Gomez, a Democratic member of the Federal Communications Commission, is asking her own agency to recognize the danger and reject the use of artificial intelligence to generate robocalls. On Wednesday, she pushed back against a petition filed by Club for Growth, a national conservative fiscal policy organization, which asked the FCC to let political groups place automated calls to people without their prior permission. The group promised limits—no more than three calls to the same person over thirty days, and an option to opt out—but Gomez was not mollified. She told Reuters that if the FCC lets a dark-money group flood Americans’ phones with AI-generated robocalls they never agreed to receive, it would bring chaos so close to Election Day. She told the Associated Press that removing the FCC’s consent requirement could produce a wave of unwanted robocalls and misinformation. The telephone is not a billboard. It is a private space, often answered with the assumption that someone real is on the other end. An AI-generated voice does not need to identify itself. It can imitate a candidate, a journalist, a neighbor, or a loved one. By the time the deception is discovered, the damage to democratic trust may already be done. Gomez’s office declined Al Jazeera’s request for further comment. But her public warning was already clear: at the moment when voters are most anxious, the last thing democracy needs is a machine that can sound exactly like a person.
The dispute before the FCC is heavily procedural, but its implications are anything but. It began with an August 31 petition from Club for Growth, a conservative group focused on fiscal policy. The organization asked the commission to allow political groups to make automated calls without first obtaining the called party’s permission—a request that would, in effect, cut the ordinary person out of the consent equation. To make the proposal look modest, the group said it would cap the calls at three per person over a thirty-day period and would allow people to opt out of future messages. The FCC opened a public comment period on September 4; the deadline for initial comments was Monday, and reply comments are due October 19. The agency has not yet ruled on the petition. This is not a brand-new question. Robocalls have been used in political campaigns for decades, and the FCC has long drawn a line that feels increasingly arbitrary: automated calls to cellphones and pagers are generally forbidden without prior express consent, but calls to landlines are still allowed. That rule was written in an older technological era, before synthetic voices could pass as human. In February 2024, the FCC moved to address the new reality, banning AI-generated voices in robocalls and saying that “this action now makes the act of using AI to generate the voice in these robocalls itself illegal.” That ban, however, does not mean the issue is settled. The Club for Growth petition asks the FCC to carve out a political exception. If it is granted, the exception could swallow the rule, because nearly every campaign will want the same permission. And once everyone has it, consent becomes meaningless. The only thing protecting the person at home will be a promise from the very people who want to call them.
The danger is not theoretical. In the run-up to the 2024 New Hampshire primary, voters received a robocall featuring an AI-generated voice in the unmistakable accent of then-President Joe Biden. The recorded message told them not to vote, suggesting they should save their strength for the November election. It was not President Biden. It was a deception built out of samples of his real voice, reassembled into a message he never said. The FCC traced the call to Steve Kramer, a political consultant who had worked for Dean Phillips, Biden’s long-shot primary challenger. The agency said Kramer could face $6 million in fines for his role. He was later acquitted of voter-suppression charges, but the incident sent a shock through election officials. If a sitting president can be impersonated so convincingly on such short notice, then no politician, no campaign worker, and no ordinary voter is safe from audio forgery. The reaction at the state level has been stronger than the federal one. Louisiana, Tennessee, and Vermont passed laws this year requiring that AI-generated campaign calls be clearly disclosed. Since 2019, twenty-five states have passed some kind of law dealing with AI-generated election materials. But these laws are patchwork. A robocall does not care about state lines. Gomez made exactly this point when she warned that removing the FCC’s consent requirement could lead to a rise in unwanted robocalls and misinformation. Misinformation, in this case, is not just a misleading flyer or a taken-out-of-context quote. It is a voice with no face, wearing the mask of a real person. When that happens in the closing days of an election, there is no time to find the truth, no courtroom to restore order, and no apology that can undo a stolen vote. The New Hampshire call was a warning shot; the FCC has been asked to decide whether the next one will be allowed.
Gomez’s warning is one strand of a larger public anxiety about artificial intelligence and the electoral process. According to a Reuters/Ipsos poll, 57 percent of registered voters said the Trump administration has not taken the risks of AI seriously enough. Fifty-four percent said the same about congressional leadership. These numbers are not fringe; they are majorities of the electorate, and they reflect something deeper. The public can see that the same technology that writes poetry and diagnoses diseases can also fabricate speeches, impersonate politicians, and flood social media with plausible lies. Yet the official responses have often been cautious to the point of toothlessness. Late last month, the White House convened leaders from some of the most powerful AI companies in the world—Anthropic, OpenAI, Meta, and Google—and persuaded them to sign a voluntary pact. In that pact, the companies agreed to put more safeguards around the development of their most advanced systems. But the accord is not binding. It does not require independent audits. It does not create fines for violations. It does not give election officials any new tools. It asks the foxes to guard the henhouse and calls that progress. Gomez’s position on robocalls is the opposite: she is asking for a binding rule, a clear prohibition, a line that does not depend on a company’s goodwill. The contrast could not be sharper. A voluntary pledge might be a meaningful first step in quieter times. But this is not a quieter time. It is an election year, the stakes are enormous, and the technology is moving faster than the rules.
The critics of voluntary AI pledges have been vocal, and their objections have weight. Tom Wheeler, a visiting fellow in governance studies with the Center for Technology Innovation at the Brookings Institution, wrote that the so-called White House Accord on Super Intelligence is “morally binding” only. That phrase sounds meaningful until you ask a simple question: who enforces morals? Wheeler noted that the accord allows AI companies to design their own controls, select their own evaluators, and determine whether their own procedures are being followed. In other words, the regulated become the regulators. He also reminded readers that history offers little reason for confidence: Facebook, Google, and Twitter have all, at various points, walked back or violated their own self-imposed standards when the pressure of competition or profit became too strong. The pattern is familiar. Companies promise to do better after a scandal, adopt a set of principles, publish a blog post, and then quietly return to business as usual when the news cycle moves on. If the most powerful tech companies in the world cannot be trusted to keep voluntary promises, why should a political group’s voluntary promise to stop at three calls be trusted? That is the uncomfortable question at the heart of the FCC petition. AI-generated robocalls are not just a consumer nuisance. They are a form of speech that can be weaponized in the final days of a campaign, when there is no time to correct the record. A voluntary promise is not a remedy. It is a delay.
The FCC has made no final decision on the petition. It is still considering the comments, still weighing the arguments, still deciding whether to allow the request, deny it, or approve it with conditions. There is still time to do the right thing, but not much. The reply comments are due October 19, and the November midterms are waiting in the wings. The commission has already heard the most important argument, and it is not in the legal briefs. It is the sound of a phone ringing at dusk. It is the hesitation of an older voter who wonders if the voice on the line is really her grandson, or a stranger with a sampler. It is the confusion of a primary voter who hears President Biden telling him not to vote and cannot quite believe his ears. The rule the FCC chooses will say something about what kind of democracy America wants. If AI-generated robocalls are allowed without consent, then political speech becomes a thing that can be manufactured in bulk, disguised as humanity, and delivered to people who never asked for it. If they are forbidden, the FCC will be saying that even in an age of synthetic voices, a citizen has a right to know who is speaking. Anna Gomez has already made her choice. She asked the agency to protect the quiet space between one person and the phone call that interrupts dinner. The rest of the commission should do the same—not because the legal arguments are complicated, but because the human ones are simple. Before the November midterms turn that space into a battlefield, the FCC should remember that an election is supposed to be a conversation among citizens, not a performance by machines.