Meanwhile, Meta said in a statement that it strongly condemns “the content that has appeared across different internet services” and has worked to remove it.
“We continue to monitor our platforms for this violating content and will take appropriate action as needed,” the company said.
A representative for Swift didn’t immediately respond to a request for comment Friday.
Allen said the researchers are 90% confident that the images were created by diffusion models, which are a type of generative artificial intelligence model that can produce new and photorealistic images from written prompts. The most widely known are Stable Diffusion, Midjourney and OpenAI’s DALL-E. Allen’s group didn’t try to determine the provenance.
OpenAI said it has safeguards in place to limit the generation of harmful content and “decline requests that ask for a public figure by name, including Taylor Swift.”
Microsoft, which offers an image-generator based partly on DALL-E, said Friday it was in the process of investigating whether its tool was misused. Much like other commercial AI services, it said it doesn’t allow “adult or non-consensual intimate content, and any repeated attempts to produce content that goes against our policies may result in loss of access to the service.”
Asked about the Swift deepfakes on “NBC Nightly News,” Microsoft CEO Satya Nadella told host Lester Holt in an interview airing Tuesday that there’s a lot still to be done in setting AI safeguards and “it behooves us to move fast on this.”
“Absolutely this is alarming and terrible, and so therefore yes, we have to act,” Nadella said.
Midjourney, OpenAI and Stable Diffusion-maker Stability AI didn’t immediately respond to requests for comment.
Federal lawmakers who’ve introduced bills to put more restrictions or criminalize deepfake porn indicated the incident shows why the U.S. needs to implement better protections.
“For years, women have been victims of non-consensual deepfakes, so what happened to Taylor Swift is more common than most people realize,” said U.S. Rep. Yvette D. Clarke, a Democrat from New York who’s introduced legislation would require creators to digitally watermark deepfake content.
“Generative-AI is helping create better deepfakes at a fraction of the cost,” Clarke said.
U.S. Rep. Joe Morelle, another New York Democrat pushing a bill that would criminalize sharing deepfake porn online, said what happened to Swift was disturbing and has become more and more pervasive across the internet.
“The images may be fake, but their impacts are very real,” Morelle said in a statement. “Deepfakes are happening every day to women everywhere in our increasingly digital world, and it’s time to put a stop to them.”