Why US lawmakers are targeting deepfakes in campaign advertising

The rise of generative artificial intelligence has reshaped how political messages reach voters, and nowhere has the alarm grown more quickly than in Washington. Synthetic media — fabricated clips of public figures and AI-cloned audio impersonating politicians — has prompted fresh calls to restrict how such content can be deployed in election campaigns.

Australia has been watching these developments closely, having wrestled with similar problems during the 2023 Voice referendum and various state elections. While Canberra has not yet proposed a federal ban of its own, regulators such as the eSafety Commissioner have fielded complaints about synthetic imagery for years, leaving Australian observers well placed to compare notes on what works.

The bipartisan push in Washington

Several bills circulating on Capitol Hill would prohibit campaigns from distributing materially deceptive AI-generated content within a set window before an election. Senators from both parties have co-sponsored measures targeting voice impersonations of candidates, while House members have drafted companion legislation aimed at visual deepfakes. The proposals vary in scope, with some applying only to federal races and others reaching down to state and local contests.

Supporters argue that the speed of generative tools has left voters with little time to verify what they are seeing during rapid news cycles. Critics counter that the rules could be weaponised to silence satire or hold campaigns liable for outside content. Either approach carries significant First Amendment implications.

Real-world incidents driving the debate

The current push did not emerge from theory alone. A high-profile incident earlier this year saw a political operative in the United States use an AI clone of a rival candidate's voice in a robocall, urging supporters to stay home during a primary. Investigations revealed the synthetic audio had been generated from a small sample of public remarks, raising alarm among officials who had previously dismissed the threat as hypothetical.

Similar tactics have appeared closer to home, with synthetic videos circulating on TikTok and X during recent Australian by-elections. The Australian Electoral Commission has noted a steady rise in reported incidents of synthetic content impersonating candidates, even if enforcement has remained patchy. The pattern has caught the attention of researchers in Melbourne and Sydney who follow disinformation trends.

How Australia has handled synthetic misinformation

Australian regulators have leaned on existing election law and the powers of the eSafety Commissioner rather than crafting purpose-built deepfake legislation. The eSafety office can order platforms to remove seriously defamatory synthetic content, and major parties agreed to internal codes of practice during the last federal cycle. Yet material from fringe accounts continues to slip through, particularly during tight contests.

Public discussion in Australia often invokes the idea of a "fair go," the expectation that a campaign should be fought on roughly equal terms. That cultural expectation has put quiet pressure on platforms to act faster when synthetic material surfaces during election cycles, even without explicit federal rules.

Free speech and the enforcement question

Civil liberties groups across the political spectrum have warned that any ban must include strong safeguards for parody, satire, and journalism. American First Amendment doctrine sets a high bar for content-based restrictions, and courts have repeatedly struck down attempts to regulate political speech. Drafters are therefore working to tailor the language narrowly, focusing on content that is both false and likely to cause confusion about when, where, or how to vote.

The alternative — a disclosure regime rather than a flat prohibition — has its own supporters, who argue that labels and watermarks offer a lighter touch. The political reality, however, is that labels rarely survive contact with a hostile actor intent on producing content designed to deceive.

Detection, watermarking, and the tech response

Major AI developers have begun embedding watermarks and signing cryptographically signed provenance data into their outputs. These technical safeguards remain unevenly adopted, and open-source tools can strip most current markers within seconds. Researchers at Australian institutions including the Australian National University have been developing their own detection methods, often funded through grants tied to election integrity.

The race between those producing synthetic material and those trying to identify it has only accelerated. For campaign staff, the calculus is shifting: a piece of content that might have circulated harmlessly a year ago can now reach millions within hours, often before a fact-checker has time to file a response.

What happens next

The legislative calendar in Washington is crowded, and election-year politics make passage far from certain before voters head to the polls. Committees in both chambers have signalled they intend to mark up versions of the bills in the coming months, with floor votes expected if leadership clears the path. Observers in Canberra, including officials at the Australian Electoral Commission, will be tracking the outcome closely, since any new US framework could shape debates at home.

Share this story with anyone tracking the synthetic media debate ahead of the next Australian federal cycle. Readers can also review the publisher's terms and conditions page for context on how stories are aggregated and verified across the site.