The US government has introduced new AI regulation that will give guidance on watermarking AI-generated content, that could theoretically lead to deepfake videos and images, including deepfake porn content, being more easily identified as AI-derived.
President Joe Biden announced a swathe of AI regulations in a new executive order that is due to be implemented over the next year. The White House said the regulations would “protect Americans from AI-enabled fraud and deception by establishing standards and best practices for detecting AI-generated content and authenticating official content.”
AI can be used to generate deepfake porn images and video: content that can appear to show a person, often a celebrity, in a fake porn situation. Some websites offer porn deepfakes of celebrities for titillation purposes, but deepfake porn can also be used for revenge porn, coercion and blackmail purposes. Deepfakes are also a significant concern in the political arena.
Some porn-generating AI services offer watermark removal or addition services, sometimes as a ‘premium’ paid-for perk.
President Biden’s new AI executive order will, according to the White House, “develop guidance for content authentication and watermarking to clearly label AI-generated content”. It said that the government doing this would show that communications received from the US government were authentic, and would “set an example for the private sector and governments around the world”.
The US government has not revealed how the watermarking plan would be implemented or enforced, or what the consequences, if any, would be for distributing non-watermarked AI content. The regulations, however, are expected to kick in soon, with President Biden saying that due to the rapidly developing nature of AI, the US government could not move at a “normal government pace”, so had pushed ahead with an executive order.
In May 2023 the Preventing Deepfakes of Intimate Images Act, which would criminalize non-consensual deepfake porn content, was proposed but is yet to pass. Sharing deepfake porn without consent is being made illegal in the UK, as part of the recently-passed Online Safety Bill.
US Vice President Kamala Harris said that there was a “moral, ethical and societal duty to make sure AI is adopted and advanced in a way that protects the public from potential harm and ensures that everyone is able to enjoy its benefits”.
Actress Scarlett Johansson has taken legal action against an AI image editor called Lisa AI: 90s Yearbook & Avatar for allegedly using her likeness in an advert without her permission.
The app reportedly released a short advert allegedly featuring an AI-generated version of Johansson’s voice and image (not, we should point out, nude or in a sexualized way). The actress’s attorney said he was handling the issue “in a legal capacity”, with her spokesperson confirming to Variety that she was not a spokesperson for the AI app.
The makers of other AI apps featuring AI clones of celebrities, that may not always be authorized, will be watching how the legal process plays out closely. Sci-match, a recently-released dating app that suggests matches based on face scans, has a feature that allows you to chat with clone versions of celebrities including actress Jennifer Lawrence.