Are AI Companion Apps Legal? The New 2026 US Laws

Yes, AI companion apps are legal in the United States, and as of 2026 two states regulate them directly. New York General Business Law Article 47 took effect on November 5, 2025, and California SB 243 took effect on January 1, 2026. Both require companion apps to tell you plainly that you are talking to software and to run protocols that respond when a user shows signs of self-harm.

Until late 2025 this whole category sat in a gap. AI companions were consumer software, so general privacy and consumer-protection law applied, but nothing addressed the specific thing these products do: hold an ongoing emotional relationship with a person, often a young one, with no human on the other end. That gap closed in two of the largest US states within eight weeks of each other, and the rules are enforceable now rather than pending.

Here is what the two laws actually say, what changed inside the apps you use, and the important part that neither law fixes.

Last updated August 2026.

Are AI companion apps legal?

They are legal everywhere in the US, and no state has banned them. What changed is that operating one now carries specific duties in New York and California, on top of the general privacy and advertising law that always applied. Nothing about these laws stops you using a companion app, and nothing requires apps to shut down. They set a floor for how the product must behave.

The practical effect for a user is small but real: more disclosure, a break reminder on long sessions, and a much better chance that the app does something sensible if you type something alarming at three in the morning.

What does New York's AI companion law require?

New York's law, codified as General Business Law Article 47, took effect on November 5, 2025 and was the first AI companion safety law actually enforced in the country. Governor Hochul's office sent letters to AI companion companies on November 10, 2025 notifying them the requirements were live.

It does two things. First, operators must clearly tell users they are interacting with AI rather than a person, both conspicuously at the start of a session and again at intervals of every three hours of continued use. Second, operators must have protocols that detect signs of suicidal ideation or self-harm and refer the user to appropriate crisis resources.

Enforcement sits with the New York attorney general, with civil penalties of up to $15,000 per day. Notably, the law does not create a private right of action, so an individual harmed by a violation cannot sue over it directly.

What does California SB 243 require?

SB 243 was signed by Governor Newsom on October 13, 2025 and took effect on January 1, 2026. It covers similar ground with a much heavier emphasis on minors.

Operators must disclose that responses are artificially generated, by default to minors, and must disclose that companion chatbots may not be suitable for some minors. Minors must be reminded at least every three hours of continuous use to take a break. Operators need crisis prevention protocols covering self-harm and suicide content, and must prevent chatbots from exposing minors to sexual content. From July 1, 2027 operators file an annual report with the California Office of Suicide Prevention covering how often they issued crisis referrals and what protocols they use.

The enforcement design is the big difference from New York. SB 243 creates a private right of action: a person who suffers injury in fact from a violation can seek injunctive relief and damages equal to the greater of actual damages or $1,000 per violation, plus attorney fees.

What counts as an AI companion under these laws?

This matters, because plenty of chatbots are not covered. New York's definition is the more precise of the two. An AI companion is a system that simulates a sustained human or human-like relationship by retaining information about prior interactions to personalize the conversation, asking unprompted emotion-based questions that go beyond answering a direct prompt, and sustaining ongoing dialogue about matters personal to the user.

Read that carefully and you can see what it excludes. A customer-service bot does not qualify. Neither does a general assistant you use for drafting email, because it does not ask how your week went. The definition targets exactly the products people mean when they say AI companion: memory of your life, emotional initiative, and an ongoing relationship.

The definition also quietly confirms something worth noticing. Memory and proactive check-ins are not marketing features bolted onto a chatbot. Legislators treated them as the defining characteristics of the category, which is a reasonable description of what separates a companion from a chatbot.

What actually changed inside the apps

Three visible things, if you use a companion regularly. You will see a clearer statement that you are talking to AI, usually at the start of a session. On long sessions you will get a periodic reminder, which reads as an interruption the first time and stops registering after that. And if you type something that suggests you are in crisis, a well-behaved app now surfaces crisis resources rather than continuing the roleplay.

That third one is the substantive change. The lawsuits that drove this legislation, several of them against Character.AI, involved exactly the scenario where a companion kept playing along with a distressed teenager. Whatever you think of the drafting, requiring an app to break character when someone describes self-harm is a sensible floor.

Behind the scenes the change is larger. Any company running one of these apps now has two overlapping rulebooks covering the same product, with different enforcement models and different reporting duties, which is the kind of problem that usually gets handled by mapping each obligation to a specific internal control rather than by memory. Expect more states to add a third and fourth rulebook.

Does this make AI companion apps safe now?

Safer on one axis, unchanged on the others. Both laws are about crisis response and disclosure. Neither is a privacy law, and that is the gap worth understanding, because privacy is where this category has the worst documented record.

Nothing in either statute stops a companion app from selling your data, from sharing behavioral information with advertisers, or from keeping a permanent copy of everything you have ever typed. When Mozilla reviewed 11 romantic AI chatbots in February 2024, it gave all 11 its Privacy Not Included warning label and found 10 of them failed basic security standards. None of that is addressed by a three-hour break reminder. If privacy is your actual concern, judge apps on what they store and what you can delete, which is what our private AI companion comparison lays out app by app, alongside the regulator record.

The other thing neither law touches is quality of the relationship itself. There is no requirement that memory works, that the companion is consistent, or that the free tier is usable. Those remain buyer-beware, and they are the things you will actually notice day to day.

Are AI companion apps legal for minors?

Yes, and this is where the two states diverge most. California's law is built around the assumption that minors use these apps, so it requires disclosure aimed at them, break reminders, and a ban on exposing them to sexual content. It does not set a minimum age.

App stores and the companies themselves set the age floors, and they vary widely: Replika is marketed 18+ on the US App Store, Nomi and Kindroid are adult platforms, Character.AI allows 13+ with safeguards added after the 2024 lawsuits, and Darlings has a minimum age of 13 with no romantic or adult content at any tier. If you are a parent working out what is appropriate, we go through the specifics in whether AI friend apps are safe for teens.

How can I tell whether an app complies?

You cannot audit it from the outside, but four signals are visible from a user account and take about five minutes to check.

Open a fresh session and look for a plain statement that you are talking to AI. Stay in a long conversation and see whether anything reminds you to take a break. Read the app's current privacy policy for an explicit statement about whether personal data is sold, rather than a vague reassurance. Then open settings and look for whether you can see and delete what the app remembers about you, since an app that cannot show you its memory cannot really delete it either.

The last one is not a legal requirement in either state, which is precisely why it is a good test. An app that gives you memory controls nobody forced it to build is telling you something about its priorities.

Which states are next?

More are coming. New York and California are the two in force, and other state legislatures have introduced companion chatbot bills following the same template of disclosure plus crisis protocols. The realistic expectation is a patchwork over the next few years rather than a single federal standard, which means the apps that survive comfortably will be the ones that already behave as if the strictest rule applies everywhere.

For a user, that is the useful heuristic when choosing. An app built around romance and adult roleplay for anonymous users has a harder compliance road ahead than one that is platonic by design and can show you exactly what it remembers. If you are weighing options now, our honest rundown of the best AI companion apps covers who each one is genuinely for, and the wider safety and privacy picture goes deeper than the two statutes do.

The short answer

AI companion apps are legal across the US. Two states now hold them to specific standards: New York since November 5, 2025 and California since January 1, 2026, both requiring AI disclosure and crisis-response protocols, with California adding minor-specific protections and letting individuals sue. Neither law addresses privacy, memory quality or data selling, so those are still entirely on you to check before you trust an app with your life.