The Ethics of Synthetic Media

The Spectrum of Synthetic Media Use

Synthetic media is not inherently good or bad — the same technology enables both beneficial and harmful applications. At one end of the spectrum, AI voice synthesis restores speech to people with ALS and other conditions that destroy the voice. Val Kilmer, who lost his voice to throat cancer, used voice cloning to speak in the 2022 documentary 'Val' and in 'Top Gun: Maverick.' AI image generation helps architects visualize buildings and medical researchers generate training data for rare diseases where real images are scarce.

At the other end, the same voice cloning technology enables phone scams. The same image generation tools produce non-consensual intimate imagery. The same text generation models create propaganda at industrial scale. The ethical challenge is not banning the technology but establishing norms and rules that preserve beneficial uses while preventing harmful ones.

Consent, Likeness, and Identity

The most urgent ethical issue in synthetic media is consent. When someone's face, voice, or likeness is used to generate content without their knowledge or permission, their identity has been appropriated. This ranges from relatively minor uses — an influencer's face used to advertise a product they did not endorse — to devastating ones, such as the creation of non-consensual intimate imagery, which overwhelmingly targets women and girls.

Existing laws were not designed for this. Right of publicity laws protect celebrities from unauthorized commercial use of their likeness, but they vary widely by jurisdiction and often do not cover non-commercial harms. Revenge porn laws in some states may cover AI-generated intimate images, but many do not. The legal framework is struggling to keep pace with a technology that makes identity theft trivial.

The ethical principle is straightforward even where the law is unclear: using someone's likeness without consent to create content they have not approved is a violation of their autonomy and dignity.

Building Norms for a Synthetic World

Every previous media technology — photography, film, Photoshop — went through a period where society had to develop norms about acceptable and unacceptable uses. Photography in the 1890s raised questions about privacy that led to Warren and Brandeis's famous 'Right to Privacy' article. Photoshop in the 1990s and 2000s led to debates about unrealistic beauty standards in advertising, eventually producing regulations requiring disclosure of retouched images in some countries.

AI-generated media is going through the same process at much greater speed. The norms that emerge will likely include: mandatory disclosure of synthetic content, consent requirements for use of real people's likenesses, provenance tracking through technical standards like C2PA, and clear legal liability for harmful synthetic content. The open question is whether these norms will develop fast enough to prevent serious damage during the transition.