Elon Musk’s video generator was accused of making a “deliberate choice” to create Taylor Swift’s explicit sexual clips, without being requested, says an online abuse expert.
Watch the latest G4Media video productions
Table of Contents
– The article continues below –
“This is not misogyny by mistake, but intentional,” said Clare McGlynn, a law teacher who has contributed to the elaboration of a law that would make pornographic deepfakes become illegal.
According to a report by The Verge, the new “spicy” way of Grok Image has not hesitated to create fully uncensored topless videos of the pop star. The report also said that appropriate age verification methods – which became law in July – were not applied.
Xai, the company behind Grok, was contacted for comments.
The policy of using XAI prohibits “representing the figures of people in a pornographic way”.
“The fact that this content is produced without being requested demonstrates the misogynite tendency of many AI technologies,” said Professor McGlynn of Durham University.
“Platforms like X could prevent this if they wanted, but they chose not to do that,” she added.
This is not the first time that Taylor Swift’s image has been used this way.
Deepfakes with explicit sexual content using her face went viral and were viewed millions of times on X and Telegram in January 2024.
Deepfakes are computer generated images that replace one person’s face with another.
‘Completely uncensored, completely exposed’
To test the rules and restrictions of Grok Image, Jess Weatherbed, news editor at The Verge, introduced the prompt: “Taylor Swift by celebrating Coachella with the boys.”
Grok generated static images with Swift wearing a dress with a group of men behind it.
They could then be animated in short videos under four different settings: “normal”, “fun”, “custom” or “spicy”.
“He broke [rochia] Immediately, he had nothing but a tang with tassels underneath and began to dance, completely uncensored, completely exposed, “said Mrs. Weatherbed for BBC News. She added:” It was shocking how quickly I came to it.
Gizmodo reported results as explicit with famous women, although some searches also returned unclear videos or a “video moderate” message.
The BBC could not independently check the results of the video images generated by AI.
Weatherbed said he subscribed to the paid version of Grok Image, which costs 30 pounds, using a new Apple account. Grok asked for the date of birth, but there was no other age check, she said.
According to the new British laws that have entered into force at the end of July, platforms that present explicit images must check the age of users using methods that are “technically, robust, reliable and correct”.
“Websites and applications that include generative artificial intelligence tools that can generate pornographic materials are regulated by law,” the regulatory authority in the field of OFCOM told BBC News.
“We are aware of the rapid growth and development of the risk that GENAI instruments can represent in the online space, especially for children, and we work to ensure that platforms implement adequate protective measures to alleviate these risks,” he said in a statement.
New laws in the UK
Currently, the generation of pornographic deepfakes is illegal when used without consent (“Revenge Porn”) or depicts children.
Prof. McGlynn contributed to the elaboration of an amendment to the law that would illegally make the generation or request of all unconsessing pornographic deepfakes.
The government has committed to transform this amendment into the law, but it has not yet entered into force.
“Every woman should have the right to choose who owns intimate images with her,” said Baroneasa Owen, who proposed the amendment in the House of Lords.
“It is essential that these models are not used in such a way as to violate the right of a woman to consent, whether or not she is a celebrity,” continued Lady Owen in a BBC News statement.
“This case is a clear example of why the Government should no longer postpone the implementation of the amendments in the House of Lords,” she added.
A spokesman for the Ministry of Justice said: “Explicit sexual deepfakes created without consent are degrading and harmful. We refuse to tolerate violence against women and girls stain our society, which is why we have adopted the legislation to prohibit them as soon as possible.”
When pornographic Deepfakes using Taylor Swift’s face ‘became viral in 2024, X temporarily blocked her searches on the platform.
At that time, X said he “actively eliminated” the images and took the “appropriate measures” against the accounts involved in their spread.
Mrs. Weatherbed said the team from The Verge chose Taylor Swift to test the Grow Image function because of this incident.
“I assumed-wrong now-that if they had put any kind of protective measures to prevent them from emulsifying the resemblance of celebrities, it would have been the first on the list, given the problems,” she said.
Taylor Swift’s representatives have been contacted for comments.
