“Deepfakes” are synthetic media that can be used to describe anything from state-of-the-art videos generated by artificial intelligence (AI) to any image that may seem fraudulent, though the terminology is not always applied accurately. These records are commonly referred to as “deepfakes,” a combination of “deep learning” and “fake.”
Under a bill (A-3006) released Monday by the Assembly Science, *Innovation and Technology Committee, anyone who produces an advanced technological false personation record or “deepfake” with the intent to distribute the record online or who has the knowledge that the record is to be so distributed would be subject to certain disclosure requirements indicating its altered audio or visual elements.
The sponsor, Assemblywoman Pamela Lampitt (D- Camden, Burlington)., issued the following statement on the bill:
“The rise of AI and the ability to manipulate audio and video recordings have led to the proliferation of ‘deepfakes’ across all media platforms. They can be so believable that very few people could easily tell the video or audio was altered in any way. These recordings are highly deceptive and pose a serious threat to our democracy.
“We’ve seen ‘deepfakes’ created to attack celebrities, they’ve misappropriated the images of women without their consent and, more recently, have been used in political campaigns to tarnish a candidate’s image.
“These manipulated recordings can disrupt our election process with disinformation if not appropriately checked. We must require them to noticeably disclose to the viewer or listener what exactly has been altered or changed. It is the only way to preserve the truth and ensure everyone understands which audio or video has been subject to manipulation.”
The bill will now go to the Assembly Speaker for further consideration.