Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Tech companies’ reliance on artificial intelligence and face-scanning tools to prevent under-18s from viewing unsuitable content will be put to the test this weekend as the UK begins enforcing one of the world’s toughest online safety regimes.
Global tech companies have raced to meet strict new government restrictions that come into effect this week preventing children from accessing sites that carry adult content, including pornography as well as other “harmful” material that relates to self-harm, eating disorders or suicide.
“There has been a rush from a number of services to get stuff in this week, and we need to work out whether that’s effective,” said Oliver Griffiths, group director for online safety at Ofcom.
Elon Musk’s X, TikTok and Reddit are among the tech groups introducing new “age assurance” systems and controls for UK users in the run-up to today’s deadline set by regulator Ofcom to begin complying with the Online Safety Act, which was passed in 2023.
Sites that are found to break the new rules, which as well as social apps include search and gaming platforms, face penalties of up to £18mn or 10 per cent of global turnover, whichever is greater.
Ofcom said it would also start monitoring how Facebook, Instagram, YouTube and other sites that say they ban adult content are preventing children from seeing harmful material.
The media regulator said it would assess these companies’ “practical actions to keep children safe” until the end of September before taking “swift enforcement action” if the tech companies fail to satisfy the law’s requirements.
Ofcom has recommended a series of what it deems to be “highly effective” tools that sites can use to verify users’ ages, including bank or credit card checks, scanning photo IDs, links to phone numbers and email addresses, and analysing facial features via photo or video.
Recommended
But many social media companies, including Meta, Google-owned YouTube and TikTok, are relying on alternative techniques that use machine-learning technology to “infer” a user’s age based on how they use the site and who they communicate with. These AI-based systems have not yet been endorsed by Ofcom, although the regulator says that its own list of proposed tools is “non-exhaustive”.
YouTube uses a “highly effective AI-driven age inference model that assesses user behaviour as part of our age verification process”, as well as offering an age-gated part of its platform.
Meta said it planned to roll out a similar AI-based system to detect when users are lying about their age in the UK in the autumn, after previously testing it in the US. It already offers “teen accounts” across its apps, including Instagram and Facebook, which offer tougher content filters and limit who can contact a young user.
Ofcom said it would begin to assess how effective these AI tools can be next week.
The UK Online Safety Act requires online platforms to adopt measures to prevent children from accessing age-inappropriate content. The legislation is among the world’s first to impose such strict controls on tech companies.
In effect this means that social media sites can either commit to removing all adult content, which is almost impossible in many cases; creating a “safe” site without adult content for under-18s; or bringing in strict new ways to verify age, from credit cards, facial age recognition or through other “highly effective” tech solutions.
“We set out earlier in the year what ‘highly effective’ means, and we’ll need to make sure that it works,” said Griffiths. “If there are failures, then people are moving on to the short list of law enforcement. We have over 6,000 websites that are committed to have highly effective age assurance by midnight [on Thursday night].”
Reddit this month began collecting and verifying the age of UK users before they can view mature content using a facial recognition software called Persona © Dado Ruvic/Reuters
Reddit, which allows its users to be anonymous, this month began collecting and verifying the age of UK users before they can view mature content using a facial recognition software called Persona using either an uploaded selfie or a photo of a government-recognised ID
TikTok said that it would introduce “additional measures” to comply with the Online Safety Act on Friday, including “enhanced deployment of age assurance technologies”.
Bluesky, Discord, Grindr and X have told Ofcom that they would use a variety of age assurance technologies on their platforms. Over the past month, adult service providers such as Porn Hub have committed to deploying age-checks, though not all have provided details on how they will do so.
Griffiths said that “there’s been a lot of eyes everywhere in terms of X . . . because there’s been a lot of research showing that children’s first access to pornography has quite often been through X”.
Critics have argued that many teenagers will turn to virtual private networks (VPNs), which disguise the location of a user so they no longer appear to be in the UK, to circumvent the new rules. Griffiths said that the real issue is “really young kids that are stumbling across porn. and they’re not out to look for it”.
But he admitted: “If you are a determined teenager looking to find porn in the same way as a [teenager] bought alcohol before 18 this is not foolproof. There are opportunities for people to use VPNs, but this is part of a broader system approach.”