The UK just became the first major Western government to regulate AI companions. It also banned social media for under-16s.
Prime Minister Keir Starmer announced the move on June 15, calling it “a line in the sand.”
The package includes a social media ban, restrictions on livestreaming and stranger contact for kids, and, buried in the small print, a minimum age of 18 for AI “romantic companion” chatbots.
What’s actually changing
Four policies, all announced together.
Social media ban for under-16s. Platforms with social interaction, user posting, and algorithmic feeds must block under-16s from accessing their services.
Livestreaming ban for under-16s. Kids can’t livestream themselves on any platform, even if the platform itself is in a different category.
Stranger communication restrictions. Methods for unknown users to contact children are restricted. The policy applies to a wider range of online services, including gaming platforms, but does not affect multiplayer gameplay itself.
AI romantic companion chatbots: minimum age 18. Services “designed to simulate sexual relationships or roleplay” must verify users are 18+. Broader AI chatbot intimate features are restricted for under-18s.
There’s also a default-on protection for 16- and 17-year-olds, meant to prevent a regulatory cliff at age 16. More details on overnight curfews and infinite-scroll breaks for under-18s are expected in July.
Which platforms are affected
The social media ban targets platforms with user-to-user social interaction and algorithmic distribution. The named list comes straight from the government press release:
In scope: Snapchat, TikTok, YouTube, Instagram, Facebook, X.
Out of scope: WhatsApp, Signal, educational services, e-commerce platforms, music streaming. Music streaming is “kept under review” but is not in the current ban.
The UK is following Australia’s December 2025 model. Australia was the first country to do an outright ban (versus age verification or parental controls). UK is the second, and the first major Western European government to follow.
The AI companion angle (the TRT hook)
This is the part most general press will underplay.
The UK has set a minimum age of 18 for AI “romantic companion” chatbots. The legal analysis from Lewis Silkin frames the scope: services “designed to simulate sexual relationships or roleplay” must age-gate to 18+.
Broader AI chatbot intimate features are restricted for under-18s. That’s a wider net than just the “romantic companion” label.
This is the first major Western government regulation specifically targeting AI companion products. The US is still debating it. France is considering it. The UK just did it.
Open question: what counts as a “romantic companion”? The policy text points to services designed to simulate sexual relationships or roleplay. Whether that covers Character.AI, Replika, Pi, and similar services depends on how Ofcom interprets the definition. The July consultation response should clarify.
When the rules kick in
The timeline is set, but it’s not fast.
Announcement: June 15, 2026.
Legislation to Parliament: Before Christmas 2026.
First regulations in force: Spring 2027. That’s roughly 9 to 12 months from announcement to enforcement.
Full consultation response: July 2026. The next 30 days will fill in the operational details.
The government is using secondary legislation under the existing Children’s Wellbeing and Schools Act. No new primary Act is needed. That means faster implementation, but it also means the policy specifics are being set via regulations rather than full parliamentary debate.
How enforcement works (or doesn’t yet)
This is the honest tension.
The government is leaning on Ofcom, the UK communications regulator, to handle enforcement. The plan:
Ofcom rapid study on “highly effective age assurance” methods. That’s the term for age verification that’s actually robust enough to be hard to bypass.
Urgent enforcement review. The Secretary of State has asked the new Ofcom Chair to review Ofcom’s enforcement capabilities and publish a strategy as soon as possible.
Funding. The government has committed to fully funding Ofcom for the new responsibilities.
The model is Australia’s, and the Australian experience has been mixed. Age verification at scale has never worked perfectly. The UK is betting that better technology and harder requirements can do what previous attempts could not.
The honest read: the government says it’ll work. The details aren’t here yet. The July consultation response will tell us more.
Public support
The political base is solid.
116,000+ responses to the government consultation. That’s one of the largest national consultations the government has run.
9 in 10 parents support a social media ban for under-16s.
Two-thirds of young people agree that under-16s shouldn’t be allowed to use at least some social media platforms.
The mandate is real. The 1 in 10 parents who disagree, and the 1 in 3 young people who don’t agree with at least some aspects, are the friction points. Implementation has to thread that needle.
The domino effect
The UK is the second country to do an outright ban. Australia was first, in December 2025.
France is considering similar legislation. Several US states have moved on age-verification requirements (though not full bans). The “domino effect” framing is overused, but here it’s earned: this is a global regulatory trend, not a one-off.
The AI chatbot restriction is the part most likely to spread first. It’s narrower than a full social media ban, and it has a clearer constituency (parents worried about AI companion products targeting kids).
Bottom line
The UK just drew a line in the sand. Whether it holds depends on whether age verification technology can actually deliver what the policy promises.
The AI chatbot restriction is the part to watch. It’s the first major Western regulation specifically targeting AI companion products. The definition of “romantic companion” matters. The enforcement timeline matters. The platforms affected matter.
More details in July when the consultation response drops. That’s when the operational specifics will become clear. For now, the policy is announced, the timeline is set, and the rest is implementation.
