Senator presses tech platforms to crack down on deepfakes before midterms

Sen. Mark Warner, Ranking Member of the Senate Select Committee on Intelligence, has sent letters to major social media companies, generative AI firms, and media software providers pressing them to move faster against deepfakes and other manipulated media ahead of the 2026 midterm elections.

He warned that more capable AI tools and weaker federal guardrails have created a dangerous opening for election-related deception.

Indeed. Since November, at least 15 campaign ads using AI-generated content have aired. Across state, local and federal races, campaigns have used AI in settings ranging from school board contests to gubernatorial campaigns.

In the Massachusetts governor’s race, the campaign of Republican primary candidate Brian Shortsleeve released an AI-generated radio ad designed to sound like Democratic Gov. Maura Healey, using a synthetic version of Healey’s voice to deliver lines she never said, including remarks about the state’s economy.

Also last week, Republican Texas Sen. John Cornyn’s campaign released an AI-generated ad attacking his primary runoff opponent, Republican state Attorney General Ken Paxton. The two will have a run-off in May. The ad parodies the song “Love Shack” by the B-52s and features Paxton in a car with two women whose faces are covered by black squares with “Mistress #1” and “Mistress #2” written on them.

Meanwhile, the Senate GOP’s official social media account posted an attack ad with a deepfake of James Talarico, the Democratic Texas state representative who will face off against either Cornyn or Paxton in November.

It is against this backdrop that Warner said maliciously manipulated media now pose a growing threat to public trust, vulnerable communities, and democratic institutions, and that the technology sector needs to adopt more concrete protections before campaign activity intensifies.

Warner’s letter amounts to a warning that the private sector has already acknowledged the deepfake threat, has already made voluntary promises, and now has only a limited window to prove those commitments mean something before manipulated content becomes an even more routine feature of the 2026 midterm landscape.

The letter is a continuation of Warner’s efforts to push tech companies to take concrete measures to combat malicious misuses of Generative AI that could impact elections.

In May 2024, he sent a letter to every signatory of the Tech Accord to Combat Deceptive Use of AI in 2024 Elections demanding specific answers about the actions companies are taking to be in compliance with this agreed upon roadmap that improves the information ecosystem surrounding elections.

“Prior to the 2024 U.S. elections, Russian-attributed actors used media manipulation techniques to denigrate a U.S. Vice-Presidential candidate, and a domestic actor utilized voice cloning software for robocalls impersonating President Biden in the New Hampshire primary,” Warner pointed out in his latest letter.

“While these malicious actions largely failed to meaningfully effect the elections, the capabilities of generative artificial intelligence products have grown tremendously in the intervening years,” Warner continued.

“Against the backdrop of an abrupt pullback in federal resources, an effective multi-stakeholder approach is needed to ensure that industry, state and local governments, and civil society adequately anticipate – and counteract – media manipulation techniques that cause harm to vulnerable communities, public trust, and democratic institutions,” Warner declared.

The Virginia senator sent his letter to a wide range of companies that together make up the modern synthetic media pipeline. The recipients included social platforms, AI developers, and editing software firms such as OpenAI, Anthropic, xAI, Meta, Adobe, ElevenLabs, Cohere, Microsoft, MidJourney, Canva, Snap, Google, Synthesia, TikTok US, Bluesky, Pinterest, and Reddit.

The breadth of that list reflected Warner’s central point that the deepfake problem no longer sits only with social networks because the tools used to generate, modify, and distribute synthetic content are now spread across the broader tech ecosystem.

The timing of the letter is also significant. Warner’s office pointed to AI-generated campaign videos released by the National Republican Senatorial Committee as a sign that synthetic political content is already becoming normalized well before voters head to the polls in November.

The concern is not simply that such content exists, but that it can circulate at high speed across platforms and be stripped of context, turning disclosure labels or campaign disclaimers into weak safeguards once clips begin traveling on their own.

Biometric Update reported last month that the danger is not simply that voters may be misled about candidates or policy positions, but that AI-generated media is increasingly positioned to suppress participation itself by instilling fear, confusion, and mistrust around the basic act of voting.

One plausible scenario election officials are quietly preparing for involves AI-generated videos depicting chaos at polling locations in heavily minority or economically disadvantaged neighborhoods, Biometric Update reported.

A short clip, circulated across social media could show armed individuals arguing outside a community center identified as a polling place, with sirens audible in the background and text warning that police have shut the site down after violence erupted.

New Jersey Democratic Sen. Andy Kim said “these deepfakes are dangerous and wrong. We need protections not just for politics, but for all Americans that could be targeted.”

Rather than making a general appeal for responsibility, Warner laid out a detailed list of measures he wants companies to adopt. For generative AI and media editing firms, he called for stronger content credentials and authenticity signals.

This includes metadata and prominent visible watermarks, along with licensing restrictions on downstream resellers, detection-sharing with trusted partners, rapid-response channels for media and civil society groups seeking authentication, and better systems for victims of impersonation campaigns to report abuse.

Warner also urged companies to proactively identify impersonation efforts using their products and to notify victims quickly.

For social media platforms and other major distributors, Warner’s demands were equally specific. He urged them to adopt and enforce clear terms of service for generative and manipulated media, consider requiring visual markers for such content, and screen uploads for content credentials, watermarks and other provenance signals.

Warner also called for internal or third-party detection tools to identify manipulated content that lacks those markers, closer collaboration with media and civil society groups, direct engagement with candidates and election officials on authentication systems, stronger victim-reporting processes, publicly accessible databases of violating synthetic media, and cross-platform information sharing related to election disinformation, voter suppression, harassment and non-consensual intimate imagery.

The letter also reflects Warner’s frustration with the gap between public commitments and operational enforcement. He pointed to the Coalition for Content Provenance and Authenticity and the Tech Accord to Combat Deceptive Use of AI in 2024 Elections as examples of industry efforts that, while useful, remain incomplete and voluntary.

The AI Elections Accord, announced at the Munich Security Conference in February 2024, committed participating companies to work together to detect and counter harmful AI-generated content aimed at elections. But the broader question since then has been whether those promises would translate into visible enforcement at the scale required for a major national election cycle.

That skepticism has been reinforced by outside reviews of the industry’s performance. A Brennan Center assessment published in February 2025 found that while major companies had publicly embraced a range of anti-deepfake measures after signing the accord, their policies often remained vague, uneven and difficult to evaluate in practice.

That critique aligns closely with Warner’s new message, which is less about persuading companies that the threat exists than about pressing them to show, in concrete terms, what they are doing to address it.

Warner tied the danger to what happened during the last election cycle. His office cited Russian-attributed efforts to manipulate media involving a U.S. vice-presidential candidate during the 2024 election period, as well as the AI voice-cloning robocalls that impersonated President Joe Biden in New Hampshire.

Warner said those incidents did not decisively alter election outcomes, but argued that synthetic media capabilities have advanced significantly since then, increasing the likelihood that similar tactics could be more disruptive in 2026, especially if platforms and toolmakers continue to respond inconsistently.

“With the prevalence of cheap or free deepfake creation tools, as well as the near total lack of moderation on said deepfakes across the social web, there will undoubtedly be an increase of negative attack deepfakes made of running candidates spread throughout the social sphere,” Ben Colman, co-founder and CEO of Reality Defender told Biometric Update last month.

The larger policy problem remains unresolved. Congress has continued to debate election-related AI measures while states move ahead with their own rules, leaving no comprehensive federal framework in place for the fast-changing synthetic media environment.

In that vacuum, Warner is using public pressure to push companies toward a more aggressive and coordinated response.

Bipartisan policymakers have begun rolling out measures to ensure that Generative AI serves the public interest, but this effort alone is not enough to stop intentional and targeted media manipulation techniques. The private sector must also proactively partner with the public sector to prevent irreparable damage to democratic elections, Warner said.

Article Topics

AI fraud  |  deepfake detection  |  deepfakes  |  digital trust  |  elections  |  generative AI  |  synthetic data  |  U.S. Government