The AI election deepfakes have arrived

Regulation

The regulation of deepfakes is a complex issue, with a patchwork of state and federal laws attempting to address the challenges posed by AI-generated images. Adi Robertson explains that while some states have enacted laws targeting non-consensual AI pornography and election-related image manipulation, there is still a lack of comprehensive federal legislation 1. The proposed "No Fakes Act" and "Defiance Act" aim to address these issues, but they raise questions about the balance between banning synthetic images and protecting free speech 2.

We start from a point of saying that you shouldn't be able to fake an image of someone without their permission, and then we're gonna create some opt-ins with some options where you're allowed to do it.

--- Adi Robertson

Additionally, the regulation of AI-generated voices, such as in robocalls, is being addressed under existing laws like the TCPA, highlighting the multifaceted nature of deepfake regulation 3.

   

Defamation

Defamation law presents a potential framework for addressing harms caused by deepfakes, but it is fraught with complexities. Adi Robertson notes that while defamation law offers a structured way to determine when a statement is an unacceptable lie, it can also be misused by powerful individuals against those unable to defend themselves 4. The idea of a federal right to likeness is controversial, as it challenges First Amendment rights and could lead to broad speech regulations 4.

I think that we're on at least more solid ground there than we are with just saying, all right, fine. You know what? Just banned deep fakes.

--- Adi Robertson

Likeness laws, primarily state-level, are being considered for federal adoption, but they risk infringing on legitimate uses like parody and satire, complicating the legal landscape further 5.

   

Watermarking

Watermarking AI-generated images is proposed as a solution to distinguish between real and synthetic content, yet it faces significant challenges. Adi Robertson discusses how watermarking could train people to expect authenticity markers, but the ease of removing watermarks undermines this approach 6. The initiative C2PA aims to establish a content provenance system with industry-wide watermarking, but its effectiveness is limited by the need for users to actively check metadata 7.

It's very, very hard to create a perfect watermarking system.

--- Adi Robertson

Despite these efforts, the widespread adoption and visibility of such systems remain hurdles, as most users are unlikely to engage with the verification process 8.