Is Character AI Legal in the UK? Online Safety Act Explained

If you've searched whether Character AI is legal in the United Kingdom, you've probably landed on a mess of contradictory headlines. In February 2026, Prime Minister Keir Starmer announced that Ofcom would be handed a broader remit to regulate AI chatbots, following high-profile cases involving minors and companion apps. That announcement kicked off a wave of speculation, but very little of it explained the actual legal position. So let's cut through it. As of now, Character AI is not banned in the UK, and using it is not a criminal offence for adults. What has changed is the regulatory pressure on the platforms themselves under the Online Safety Act 2023, plus the government's stated intention to close gaps that specifically involve generative AI companions. This guide walks through what the law currently covers, where the grey areas sit, what Ofcom is actually empowered to do, and how the rules land differently depending on whether you're an adult user, a parent, or a teenager. If you're weighing up Character AI or considering alternatives, you'll finish this with a clear picture of your legal footing.

The short answer: legal to use, increasingly regulated to operate

Character AI is legal to access and use in the United Kingdom for adults. There is no UK statute that criminalises signing up to a chatbot service, having conversations with an AI persona, or generating fictional dialogue — including romantic or dramatic content — provided the outputs themselves don't fall foul of existing laws around illegal content (for example, material depicting child sexual abuse, incitement to terrorism, or content that breaches the Malicious Communications Act).

What's shifting is not user legality but platform obligation. The Online Safety Act 2023 came fully into force during 2025, with Ofcom as the designated regulator. It imposes duties of care on services that host user-generated content or facilitate user-to-user communication. Character AI, because it lets users create and share character personas that other users interact with, sits squarely within scope. That means the company — not you — carries the legal risk of non-compliance, including fines of up to £18 million or 10% of global turnover.

Starmer's February 2026 announcement extended this direction of travel. Rather than passing a new Act, the government signalled it would use existing powers plus secondary legislation to formally bring generative AI companions under Ofcom's supervisory umbrella. Expect codes of practice around age assurance, crisis-response protocols (particularly around self-harm and suicide), and transparency requirements about how models are trained and moderated. None of this makes the service illegal for you to use. It makes running the service harder, which is why some platforms have already geo-blocked UK users or pulled certain features rather than comply.

What the Online Safety Act actually requires

The Online Safety Act works through categories of harm and categories of service. For chatbot platforms, three duties matter most.

First, illegal content duties. Platforms must proactively prevent users from encountering priority illegal content. For an AI companion service, that means the model shouldn't be producing CSAM, terrorist content, or content facilitating serious violence — and if users try to steer it there, the platform must intervene. This applies whether the output is user-authored or generated by the AI itself, which is a meaningful expansion of previous UK law.

Second, child safety duties. Any service likely to be accessed by children (a low bar — essentially any consumer service without robust age gates) must assess risks to minors and mitigate them. This is where age assurance comes in. From mid-2025, Ofcom's guidance made it clear that self-declared age tick-boxes don't cut it. Services offering pornographic content or high-risk interactions to adults must use highly effective age verification — photo ID checks, credit card verification, or facial age estimation.

Third, transparency and user empowerment. Category 1 services (the largest) must publish transparency reports, offer adult users tools to filter certain content, and provide accessible complaints routes. Character AI's UK user base likely puts it in scope for a version of these obligations even if not full Category 1 status.

What the Act does not do: it does not ban AI companions, does not criminalise adult users, and does not restrict fictional or roleplay content between consenting adults. The line runs at illegal content, protection of minors, and platform accountability — not at moral disapproval of the format itself.

Why Starmer's announcement changed the tone, not the law

The February 2026 announcement was politically significant but legally incremental. It was triggered by a series of stories — including a coroner's inquest that examined a teenager's interactions with a chatbot in the weeks before their death — that made AI companions a mainstream policy concern rather than a niche tech issue.

What Starmer announced was essentially threefold: a formal Ofcom remit extension to cover generative AI chatbots as a distinct category; a consultation on mandatory crisis-intervention design standards (so that mentions of suicide or self-harm trigger clearly signposted human-support routes rather than continued roleplay); and a review of whether the Age Appropriate Design Code needs updating to reflect AI-native risks like sycophancy, parasocial attachment, and simulated intimacy with minors.

None of this changes the position for an adult UK user today. You can still open Character AI, sign up, and use it. What you may notice over the next 12–18 months is stricter age verification at signup, more assertive content interventions in certain conversation topics, and possibly reduced feature availability if the platform decides UK compliance costs outweigh the market opportunity. Some smaller competitors have already made that calculation. Larger ones, including Character AI itself, have generally chosen to comply rather than exit.

If you're an adult who wants a companion platform that has taken UK compliance seriously and offers robust safety design, mainstream options like Replika and Nomi AI have publicly leaned into safety-by-design. If you want an uncensored experience, you'll need to look at platforms that either operate offshore or use age-gated adult tiers — and understand that the legal responsibility for what you generate still rests partly with you.

Adults vs. minors: where the rules diverge sharply

For adults, the practical legal position is straightforward. You can use Character AI or any alternative. You can generate fictional content, including romantic and sexual roleplay if the platform permits it and you've passed age verification. You cannot lawfully generate content that would be illegal offline — that includes any sexualised depiction of a minor (real or fictional), content that harasses an identifiable real person in ways that meet the Malicious Communications Act threshold, or content assisting serious crime. The AI doesn't create a legal shield: if you prompt a model to produce illegal material, the material is still illegal.

For minors, the framework is much tighter and getting tighter. Under 18s should not be accessing adult-tier features on any compliant platform. Under 13s aren't supposed to be signing up at all under most platforms' terms and under UK data protection rules. The Age Appropriate Design Code, enforced by the ICO, requires services likely to be used by children to default to high-privacy settings, avoid nudging behaviours that extend engagement, and not profile children for content recommendation in ways that exploit vulnerability.

The practical challenge is that age assurance across the industry is inconsistent. A teenager who lies about their age on signup is technically breaching terms of service and, on some platforms, misrepresenting age to access age-restricted material — but the primary legal accountability falls on the platform for having weak checks, not on the child.

Parents concerned about a teenager's use of Character AI or similar apps have practical options: device-level content filtering, honest conversations about parasocial attachment, and steering toward platforms with clearer minor-safe modes. Blanket bans tend to push use underground rather than resolve the underlying pull of the format.

What could actually get you in legal trouble

The realistic risks for an adult UK user are narrow but worth naming clearly.

Generating or requesting sexualised content involving minors — even fictional, even cartoon, even AI-generated — is a serious criminal offence under the Protection of Children Act 1978 and the Coroners and Justice Act 2009. This is the single most important line. It is not a grey area. It applies regardless of how the content was produced.

Using an AI persona to harass, threaten, or defame a real, identifiable person can trigger liability under the Malicious Communications Act 1988, the Communications Act 2003, or civil defamation law. Sharing generated intimate images of real people without consent is now a specific offence under recent reforms to intimate image laws.

Using chatbots to obtain operational information to commit terrorism, serious violence, or fraud carries the same criminal exposure as obtaining it any other way. The model being the source doesn't launder the intent.

Outside these areas, adult use of Character AI and its competitors is a normal consumer activity. If you're comparing platforms and want to see how different services handle UK compliance and content policy, browsing a curated set of Character AI alternatives can shortcut the research — though be aware that the compliance landscape is moving quickly and any comparison you read may already be slightly out of date.

What to expect over the next 12 months

Three developments are worth watching if you're a UK user of AI companions.

First, age verification will get more assertive. Expect Character AI and most mainstream competitors to require some form of verified age check to access anything beyond a heavily filtered baseline. This may be facial age estimation, ID upload, or credit card verification. Privacy-conscious users won't love it, but Ofcom has been explicit that self-declaration is no longer acceptable.

Second, crisis-response design will become standardised. If you type anything the model reads as a suicide or self-harm signal, expect immediate, non-negotiable signposting to Samaritans or NHS 111, potentially with the roleplay session paused. This isn't optional design; it's likely to become a codified duty within the next Ofcom code of practice.

Third, some platforms will exit the UK market rather than comply. Smaller uncensored services in particular may geo-block UK IPs. Using a VPN to circumvent this isn't a criminal offence for the user under current law, but it does breach most platforms' terms of service and can void any consumer protection you'd otherwise have. It also means you're accessing a service that has deliberately chosen not to meet UK safety standards, which shifts more of the practical risk back to you.

The overall direction is not prohibition. It's a UK market where AI companions remain available to adults but where platforms have to earn the right to operate through demonstrable safety design. For most users, that's a net positive.

Frequently asked