On January 1, 2026, California became the first state to put companion chatbots under a dedicated safety law, and AI companion age verification stopped being a design preference and became a compliance obligation. The gap between what the law asks for and what most apps actually do is wider than the headlines suggest. In practice, the majority of age gates on companion software still verify one thing: that a user was willing to click a button.
The law asks operators to know, not to guess
California SB 243, signed in October 2025 and effective at the start of this year, is built around a conditional. Most of its strongest protections apply to users the operator “knows” to be minors.
For those users, the operator has to disclose that the conversation is with AI, and has to interrupt with a break reminder at least every three hours of continuing interaction. Operators also have to maintain and publish a crisis protocol covering self-harm content, and beginning July 1, 2027, report annually to California’s Office of Suicide Prevention on how often crisis referrals were issued.
The conditional is doing enormous work. An operator that never establishes age has fewer users it “knows” to be minors, and therefore fewer obligations triggered. That is the structural weakness critics keep pointing at, and it is why the phrase age assurance has started replacing age verification in policy conversations.
Self-attestation is a door, not a lock
Most companion apps satisfy age requirements the cheapest way available: a date-of-birth field at signup, sometimes a single “I am 18 or older” checkbox. This is self-attestation. It creates a record that the user claimed to be an adult, which has real legal value for the operator, and close to zero verification value.
The stronger methods all cost something. Document upload means collecting a government ID. Credit card checks exclude adults without cards. Facial age estimation means running biometrics on every new user. Device-level or app-store signals are less invasive but depend on the platform, and on a parent having set the device up honestly in the first place.
None of these is free, and each pushes the operator further into holding sensitive data about people who came for a conversation. That tension is the reason the industry has moved slowly, and it is inseparable from how private companion conversations really are once an app starts holding identity documents alongside chat logs.
Why several apps skipped the gate and closed the door
The most consequential response to 2025’s regulatory pressure was not better verification. It was exclusion.
Character.AI announced in October 2025 that under-18 users would lose open-ended chat access entirely, with the change taking effect on November 25. The company redirected younger users toward a more constrained creative product built around stories and generated scenes rather than free-form back and forth. The decision followed lawsuits, a Bureau of Investigative Journalism investigation, and an FTC 6(b) inquiry into how several AI companies assess mental health risk for minors.
Read plainly, that is a company deciding that verifying age well enough to serve teens safely was harder than not serving them. Several smaller operators reached the same conclusion and repositioned as adults-only. The practical result is that a growing share of companion software is now explicitly 18+, and the age gate is less about routing minors to a safer tier than about keeping them out.
The demand did not disappear when the gate went up
A Common Sense Media survey of 1,060 US teens, fielded in April and May 2025, found that nearly three in four had used an AI companion, and about a third had chosen one over a person for a serious conversation. Half said they distrust advice from AI, and eighty percent said real friendships matter more, which complicates the simplest version of the alarm.
The number that matters for age verification is different: younger teens in that sample trusted AI companions notably more than older teens did. Whatever a gate accomplishes, it does not touch that literacy gap. A thirteen-year-old who enters a false birth year is not defeating a security system. They are answering a question, and the answer is unverified by design.
This is the part of the policy conversation that tends to get skipped. Age verification allocates liability efficiently. It changes behavior weakly.
What a gate cannot reach
An age check happens once, at the threshold. Nearly everything people worry about happens later, in the conversation, and is a property of how the software is designed rather than who was let in.
SB 243 acknowledges this in places. It bars companion chatbots from claiming to be a licensed professional such as a therapist or physician, which is a design constraint rather than an access constraint, and a useful one given how often users test what an AI companion can and cannot do in a difficult moment. It requires disclosure when a reasonable person could be misled into thinking the system is human. It requires a crisis protocol.
Those are the provisions that operate on every user, verified or not. They are also the ones that are hardest to audit from the outside, because compliance lives in model behavior rather than in a signup form. An app can pass every age check and still be built to make leaving feel costly.
What this means if you are an adult using companion software
The honest summary is that the age gate on the app you use probably tells you very little about how carefully the rest of it was built. It is worth reading past it.
Ask whether the app publishes a crisis protocol, since that is now required of covered operators and is a low-effort thing to check. Ask what happens to your data if you upload an ID, and whether identity verification is stored separately from conversation history. Notice whether the product ever tells you it is AI without being asked, and whether it discourages you from ending a session.
Vinfluencer is an AI conversational companion application built for adults, and we treat the disclosure and crisis provisions of the new rules as a floor rather than a target. The age gate is the least interesting safety feature any companion app has. What the software does in the thousandth message matters considerably more than what it asked in the first one.
Frequently asked questions
Is AI companion age verification legally required in the United States?
Not nationally. California’s SB 243 requires covered companion chatbot operators to establish age verification mechanisms and took effect January 1, 2026. Several other states have introduced similar bills. Federal proposals exist but there is no single national requirement as of September 2026.
Does self-attestation count as age verification under SB 243?
The law does not enumerate one approved method, and most operators currently rely on self-attestation at account setup. Whether that satisfies a regulator in an enforcement action has not been tested. Stronger age assurance methods include document checks, credit card verification, and facial age estimation.
Are AI companion apps now 18+ across the board?
No, but the direction is clear. Character.AI ended open-ended chat for under-18 users in November 2025, and several other operators repositioned as adults-only rather than build compliant minor experiences. Some apps still offer restricted teen tiers.
Does age verification make an AI companion app safe?
Age verification controls who enters, not what happens afterward. Disclosure requirements, crisis protocols, restrictions on claiming professional credentials, and the design of the conversation itself do far more to determine whether an app behaves responsibly toward the person using it.