What is the UK's new deepfake detection plan, and how will it work?
By producer James Hockaday
The government has said it will work with tech giants, including Microsoft, to put together what it calls a "world-first" deepfake detection system.
Experts and academics will help shape a plan to tackle the growing issue of deepfakes, which in some cases can be used to create intimate images of people without their consent or facilitate other crimes.
“The devastation of being deepfaked without consent or knowledge is unmatched, and I have experienced it first hand,” said safeguarding minister Jess Phillips.
“For the first time, this framework will take the injustice faced by millions to seek out the tactics of vile criminals, and close loopholes to stop them in their tracks so they have nowhere to hide.
“Ultimately, it is time to hold the technology industry to account, and protect our public, who should not be living in fear.”
What is a deepfake?
Deepfakes are computer generated videos, audio or pictures, generated using artificial intelligence (AI) that portray events or things that do not exist in reality.
The term uses "deep" as a reference to deep-learning, a type of machine learning used by many AI models, while the "fake" part of the word is self-explanatory.
How much of a problem are deepfakes in the UK?
"Every person in the UK faces growing risk from harmful deepfakes," the Home Office says, adding that many of this media is "designed to deceive".
"These materials are often being used by criminals to deceptively steal money, strip away the dignity of women and girls, and spread harmful content online," the department said.
"Criminals are already using this technology to impersonate celebrities, family members and trusted political figures in sophisticated scams, with tools to create convincing fake content being cheaper and more widely available than ever before, and little to no technical expertise required."
This technology "does not discriminate", says Phillips, who gives examples of "a grandmother deceived by a fake video of her grandchild, a young woman whose image is manipulated without consent and a business defrauded by criminals impersonating executives".
The government estimated up to eight million deepfakes were shared in 2025 alone, up from 500,000 in 2023.
Research by the Alan Turing Institute in 2024 suggested 15% of adults in the UK reported being exposed to harmful deepfakes, including deepfake pornography, deepfake fraud, and other potentially harmful content such as content used to spread misinformation or propaganda.
More disturbingly, research published by Girlguiding found that 26% of children have seen a sexualised deepfake of a celebrity, friend, teacher or themselves.
In January the UK's data regulator Ofcom launched an investigation into the social media platform X and its sister company xAI, after the chatbot Grok was used to generate sexualised deepfake images without consent.
The Centre for Countering Digital Hate (CCDH) found that Grok had been used to produce an estimated three million sexualised images including 23,000 of children over the space of just 11 days.
X is also facing investigation by the European Commission, and on Tuesday its offices in Paris were raided by French prosecutors as part of a preliminary investigation into alleged offences including spreading child sexual abuse images and deepfakes.
X has said it has "zero tolerance for any forms of child sexual exploitation, non-consensual nudity, and unwanted sexual content" and takes action to remove such content. It has also restricted Grok's image creation and image editing abilities to paid subscribers only in a bid to improve accountability.
Cybersecurity expert and Fellow at the Institution of Engineering and Technology Dr Junade Ali told ITV News: "The true scale of the issues caused by deepfakes is not yet understood.
"But they are clearly already having a real-world impact; including misinformation, scam calls and in the criminal justice system."
Alan Woodward, professor of cybersecurity at the University of Surrey, said that "sadly", the problem is "getting bigger".
"Deepfakes being posted was estimated in the hundreds of thousands in 2023 and later in 2025 that had reached tens of millions," he said. "With the latest AI tools it’s difficult nut to conclude that the number is not getting larger faster."
How could the plan work?
The new framework being worked on will explore how technology can recognise and assess deepfakes and “set clear expectations” for industry detection standards, the Home Office has said.
Collaboration between the government, technology firms, academics and experts will underpin the development of the plan.
We don't know exactly what the finished system will look like, as applications to take part in this project were opened just before Christmas last year.
However, Dr Ali says that there are "broadly two distinct strategies for detecting deep fakes".
"One is reviewing the material to check if it is likely to be produced using AI (often using other AI models), the other is to check the provenance to see if it comes from a reputable source," he says.
"It is vital there is development in both approaches to get ahead of this issue as AI deepfakes are getting harder to spot."
Professor Woodward added: "No system will ever be 100% effective but I think it’s going to a combination of technologies from analysing video for micro expressions through to digital watermarking.
"There’s a whole host I know are being examined. Of course, this has to be used widely, regulated for and followed up. A combination of tools and legislation."
He says the Home Office is already building tools and datasets to test these tools as they're made, and that Microsoft has "already developed some powerful and widely used tech for image analysis", such as PhotoDNA – making the company well placed to help.
How to spot a deepfake
"Often the easiest way to identify if something is suspicious is to check it came from where it claims to come from," says Dr Ali.
"For example; consider hanging up on a call, and then calling the individual or organisation back on a number you know to be theirs.
"For businesses, security in a world of AI deepfakes means moving to using secure web portals which require logging-in instead of relying on phone calls and paper documents."
Subscribe free to our weekly newsletter for exclusive and original coverage from ITV News. Direct to your inbox every Friday morning.
Professor Woodward says the question of how to protect yourself from deepfakes is "a really tricky" one.
"I could give really obvious answers such as being careful about what you post, for example, nothing that is easily manipulated to bring a deepfake. But it’s scary how good the tech is so we’re in the classic whack-a-mole situation.
"As well as watching out for misuse of your image for unsavoury pictures, the main risk is ID fraud. Try to use systems that have a degree of liveness checking and not just simple facial recognition.
"Most of the big vendors do this but a little homework goes a long way."
Want a quick and expert briefing on the biggest news stories? Listen to our latest podcasts to find out What You Need To Know…