A wide range of leading technology companies have signed a voluntary agreement to combat AI-generated deepfakes that could be used for election interference by misleading voters.
The agreement proposes that signatories will adopt a new framework to help identify deceptive deepfakes and label them as they are created or uploaded to social media.
Among those who signed were Meta, Google, Microsoft, IBM, Amazon and Adobe, while social media platforms X, Snap and TikTok also signed the agreement.
Deepfakes are expected to pose a major threat to this year’s elections
AI was a key focus at this year’s Munich Security Conference, where the agreement was signed, with Google launching its AI Cyber Defense Initiative just days before.
The latest agreement will help boost communication about when and where political deepfakes are being created and distributed to limit the potential harm they cause. Although the agreement is voluntary, deepfakes are a major cause for concern in a year of critical elections around the world.
In a press release Announcing the agreement, Microsoft Vice Chairman and President Brad Smith said: “The challenges are formidable and our expectations must be realistic… There is no way the technology sector can on its own protect elections from this new type of election abuse. ”
Shortly after the Russian invasion of Ukraine, a deepfake video from Ukrainian President Volodymyr Zelensky was widely shared, in which the president asked his troops to lay down their arms and surrender. The video was quickly debunked as a fake, but deepfaking technology has made significant progress in recent years.
Just last month, the US got its first taste of election interference in 2024, when New Hampshire was targeted by a robo-call impersonation of President Joe Biden, who asked voters not to vote in the state’s primaries.
Through TechCrunch