The UK government has announced a collaboration with Microsoft and top academics to build a robust defence against the skyrocketing threat of deepfakes.
This new initiative centres on developing a standardized evaluation framework designed to identify critical gaps in deepfake detection.
By testing current technologies against real-world threats – including fraud, impersonation, and non-consensual sexual abuse – the government aims to establish clear benchmarks for the tech industry to meet.
The urgency of the project is underscored by staggering growth in synthetic media. Official figures reveal that an estimated eight million deepfakes were shared in 2025 alone, a massive jump from just 500,000 two years prior.
Criminals are increasingly using these AI-generated images and audio to defraud the public, often targeting vulnerable individuals with sophisticated scams.
Beyond individual fraud, the initiative seeks to protect national security and public trust. Last week, the Home Office funded a “Deepfake Detection Challenge” hosted by Microsoft, where over 350 experts from INTERPOL and the “Five Eyes” intelligence community were tasked with identifying manipulated media in high-pressure scenarios involving election security and organized crime.
“Deepfakes are being weaponised by criminals to defraud the public, exploit women and girls, and undermine trust in what we see and hear,” says Tech Secretary Liz Kendall. “The UK is leading the global fight against deepfake abuse, and those who seek to deceive and harm others will have nowhere to hide.”
Consumer advocates have welcomed the move but are calling for faster regulatory enforcement to protect people from financial ruin. Adds Rocio Concha, Which? Director of Policy and Advocacy:
“The UK is in the grips of a scam epidemic – social media platforms are littered with convincing deepfakes designed to con people into parting with their hard-earned cash.
“Under the Online Safety Act, platforms have duties to detect and remove fraudulent content, including deepfake scams, and the government’s plan to develop a standard for identifying deepfakes could help them do this.
“For this new initiative to work, Ofcom should not hesitate to take action – including robust fines – against companies who aren’t playing their part. Many deepfakes feature in paid-for scam ads.”
The framework is part of a broader legislative push that includes criminalizing the creation of non-consensual intimate deepfakes and banning the “nudification” tools that facilitate such abuse.
Credit: Photoillustration by Kaley Mantz/Fresh Take Florida
A sweeping new law in Florida that took effect Wednesday makes it illegal to produce sexual images of a person using artificial intelligence or similar technologies without their permission.
The new law also allows people whose photographs were manipulated that way to sue those responsible in civil court.
The law took effect this week only two days after Marion County sheriff’s deputies arrested Lucius William Martin, 39, of Eustis, Florida, and accused him of using AI to produce nude images of the juvenile daughter of someone close to him and her friend. The software Martin used digitally removed the girls’ clothing in pictures he downloaded from social media, according to court records.
Such tools can be used to “nudify” an otherwise innocent photograph.
Martin was arrested Monday and remains in the county jail in Ocala, facing eight felony counts of child pornography under Florida’s existing statutes and one count of trying to destroy evidence. The girl’s mother captured a screenshot of the images to give to authorities, the sheriff’s office said. A deputy said Martin reset his phone as he was being arrested to delete the evidence.
Martin couldn’t be reached immediately for comment because he was still in jail. He was being appointed a public defender on Thursday for his arraignment scheduled next month, but no lawyer had yet been assigned to represent him.
The versions of the images of the girls nude on Martin’s phone included remnants of their clothing that had been digitally removed and showed deformities on the girls’ arms and legs, which a deputy wrote in court records “is common on AI-generated imagery.” His phone also contained the same, unaltered images of the girls wearing clothes, court records said.
Last year, singer Taylor Swift was the victim of AI-generated, fake images of her nude, also called “deepfakes,” circulating over popular social media sites.
The Florida bill, sponsored by Republican Reps. Mike Redondo of Miami and Jennifer Kincart Jonsson of Bartow and known as the “sexual images” bill, passed the Legislature unanimously earlier this year and was signed into law by Gov. Ron DeSantis in May.
Rep. Michelle Salzman, R-Cantonment, said during a House Judiciary Committee hearing earlier this year that her community in Florida’s Panhandle has suffered cases of AI-generated sexual images.
“Seeing this brought forward is a breath of fresh air,” she said. “AI is incredible. We need it. It does a lot of good, but with great power comes great responsibility, and a lot of folks aren’t taking responsibility for their actions.”
Key provisions of the new law include criminalizing use of AI to generate a nude image of an actual person without their consent, or soliciting or possessing such images. The new felony punishment includes a prison term up to five years for each image and a fine up to $5,000.
The new law was long overdue, said former Sen. Lauren Book, a leading advocate for sex crime victims. She said AI and popular software tools make it easy to create realistic images.
“Legislation is a crucial step in ensuring that our justice system can keep pace with technological advancements so that we are not lagging in protecting our children,” said Book, a child sex abuse survivor who founded Lauren’s Kids, a nonprofit dedicated to stopping child sex abuse.
Such digitally altered images of children or teens are often used to extort families, said Fallon McNulty, executive director at the National Center for Missing and Exploited Children. Criminals can extract payment or sexual favors in exchange for agreeing not to distribute nude images to victims’ friends, classmates or family members.
The center’s CyberTipline, which started tracking reports involving generative AI in 2023, received 4,700 reports involving AI-generated images in its first year. In the first six months of 2025, she said the tipline had received 400,000 such reports.
McNulty said mainstream software companies try to block and report illicit use of their programs, but some developers offer apps with no built-in safety measures.
Meta announced earlier this year it was suing a company in Hong Kong that it said ran ads on its platforms to promote an app that helps users create nonconsensual, sexualized images using AI. It sued the developer of an app called CrushAI, which could be used to create nude images.
Lawmakers are always “trying to play catch-up” when it comes to regulating AI, said Elizabeth Rasnick, an assistant professor at the Center for Cybersecurity at the University of West Florida, adding that they are “doing the best they can with what they currently have.”
“ There’s no possible way we can foresee how these tools are going to be used in the future,” Rasnick said. “The Legislature is always going to have to try to fill in whatever gaps there were after those gaps are discovered and exploited.”
Digitally altering images has been possible for decades using specialized image-editing tools, but the new AI programs can turn out sexual content in seconds with no special skills required, said Kevin Butler, a professor of computer science and director of the Institute for Cybersecurity Research at the University of Florida.
Using the new AI tools can take a photo posted on social media and “undress the whole family,” said Kyle Glen, commander of the Central Florida Internet Crimes Against Children Task Force. He praised the new law but noted that juvenile offenders — who may try to bully classmates by creating such images — often aren’t prosecuted criminally the first time they are caught.
“As much laws as we pass and as much software is out there, and technology that we use, bad guys are always a step ahead,” Glen said. “They’re innovative and they’re going to think of ways to get around law enforcement or exploit children, you know, if that’s what they’re infatuated with.”
___
This story was produced by Fresh Take Florida, a news service of the University of Florida College of Journalism and Communications. The reporter can be reached at maria.avlonitis@freshtakeflorida.com. You can donate to support our students here.
