Tech giants sign pact to fight AI election deepfakes

Last month, voters in New Hampshire received a robocall from President Joe Biden telling them not to vote in the primary. It sounded remarkably real, but the voice was entirely fake, whipped up by an AI voice clone in seconds.

With over 50 countries holding national elections in 2024, the biggest tech players in the world are finally teaming up to contain the fallout.

At the Munich Security Conference, 20 major tech firms, including OpenAI, Meta, Google, Microsoft, Adobe, Amazon, TikTok, Stability AI, and Elon Musk’s X, signed a joint accord to combat deceptive AI deepfakes aimed at voters.

Detecting and labeling, not outright banning

The pact calls on companies to voluntarily adopt reasonable precautions to prevent AI from messing with democratic elections. The focus is specifically on deceptive synthetic media targeting political candidates, election officials, and voting procedures.

Under the framework, the companies agreed to develop detection tech, add clear labels to AI content, evaluate their own models for misuse, and share best practices across the industry.

Crucially, the companies did not commit to banning or removing deepfakes across the board.

Instead, the accord carves out protections for context, noting that educational, documentary, satirical, and legitimate political speech must still be protected.

Meta’s president of global affairs, Nick Clegg, basically admitted that heavy-handed moderation won't cut it. He pointed out that no one in the tech industry believes playing whack-a-mole or forcing a straitjacket on every platform is going to solve a massive technological shift.

Polite promises versus enforcement

Not everyone is convinced that a voluntary handshake will keep election chaos at bay.

Advocates quickly noted the absence of binding enforcement. Rachel Orey from the Bipartisan Policy Center pointed out that the language is softer than hoped, and while companies have skin in the game, the lack of real teeth means everyone has to watch whether they actually follow through.

Meanwhile, European Commission Vice President Vera Jourova welcomed the positive framework but warned that AI-driven disinformation could spell the end of democracy if left unchecked.

The scramble is very real. Days before Slovakia's election last November, a fake audio tape of a candidate allegedly plotting to rig the election and raise beer prices flooded social feeds before fact-checkers could keep up. With the U.S. Congress still dragging its feet on AI legislation, tech platforms are effectively writing their own rules.

Why it matters

For everyday voters, do not assume this pact will keep your social feeds clean. Most of the watermarking and detection systems promised by these platforms haven't even rolled out at scale yet.

In our view, this accord is a savvy defensive play by Big Tech. By agreeing to voluntary detection and labeling, the companies look proactive while carefully avoiding the massive legal and political landmine of deciding what content to censor or delete.

The obvious catch is that self-regulation rarely holds up when the pressure mounts. With companies as culturally distinct as Meta and X sharing the same table, their standards for what counts as deceptive or satirical are guaranteed to clash. When a high-stakes deepfake drops hours before a vote, polite promises and best practices might not be enough to stop the damage.