Companion Chatbot Laws by State: What NY and CA Require (2026)

Short Answer

Two state laws on AI companion chatbots are already in force: New York's, which took effect on November 5, 2025, and California's SB 243, which took effect on January 1, 2026. Both require the chatbot to say it is not a person and require the operator to run a protocol for suicide and self-harm. They differ on who gets the protections, how they are enforced and what a company risks. Many more state laws are reported to start in 2027, so which law applies to you depends on where you live and when you are reading this.

This page sets out what the two laws that are already in force require, what the other states are reported to be doing and where the reporting disagrees, and how to read a companion-bot notice you see on screen. We read the sources named below and wrote this ourselves; this is a legal-landscape summary, not legal advice, and we did not test any companion product for this page.

What counts as a companion chatbot under these laws

The laws are aimed at chatbots built to hold a relationship-style conversation, not at every chatbot. A Morrison & Foerster client alert on the two laws says California's SB 243 covers a system that gives "adaptive, human-like responses to user inputs" and can "meet a user's social needs." New York's Article 47 of the General Business Law covers an AI companion that "simulates a sustained human or human-like relationship with a user," and the same alert lists three tests: the system retains information from earlier interactions, asks unprompted questions about emotions, and keeps up an ongoing personal dialogue.

That definition matters because the same company can ship both kinds of product. A customer-service bot that answers billing questions is a different thing from an app whose pitch is a friend who remembers your week. Tech Policy Press, which counted twelve state laws in this area, reports that most of them exempt business-use bots and narrow features. If you are trying to work out whether a law applies to a product, the question is not whether it uses a language model but whether it is built to sustain a social or emotional relationship.

New York and California side by side

Source for the table: the Morrison & Foerster alert linked in the sources below, which summarizes the statutes rather than reproducing them. We have not been able to read the enrolled California bill text directly from the legislature's site in this review, so check the statute itself before relying on a figure. One law-firm summary we saw dates SB 243 as effective July 1, 2027. The likelier explanation is that this date belongs to the reporting duty, which matches the alert, but that is our reading, not something we could confirm.

What the difference means in practice

Two design choices stand out when the laws are read together, and both change what a user can actually do about a violation.

• Who gets the notice. New York's repeated notice protects every user, adults included. California ties its break reminder to users the operator knows are minors, which makes the protection depend on what the operator knows. A company that never asks for an age has less to act on.
• Who can enforce. New York leaves enforcement to the Attorney General. California adds a private right of action, so a harmed user or a parent could in principle bring a claim themselves. Whether that works in practice depends on proof, and no court outcome under SB 243 is something we are reporting here.
• What the law does not say. The quoted requirements are about notices and protocols. Neither summary describes a rule that bans emotional dependency, caps session length for adults or requires a company to keep a companion character stable. A notice every three hours is a reminder, not a limit.

The other states, and how far to trust the lists

Tech Policy Press counts twelve state laws on companion bots: New York, California and Hawaii as already in effect, and Colorado, Connecticut, Georgia, Iowa, Idaho, Nebraska, Oregon, Rhode Island and Washington listed for 2027. It reports that all twelve require disclosure that the user is talking to an AI, that all twelve require detection of suicidal ideation or self-harm with referral to resources, and that every one except New York and Rhode Island adds protections for minors. It also reports that Colorado's attorney general released proposed rules on August 11, with public comments due October 26.

A law-firm overview we read adds some dates. It reports that Oregon's SB 1546 was signed in March 2026 and takes effect on January 1, 2027, that Nebraska's LB 525 (the Conversational AI Safety Act) was enacted on April 14, 2026 and takes effect on July 1, 2027, and that Tennessee's SB 1580, which bars AI from presenting itself as a licensed mental-health professional, takes effect on July 1, 2026. The same overview gives the Washington bill number in a way that matches Oregon's, which is a sign of a copy error somewhere in the secondary reporting.

So treat every non-New-York, non-California entry in these lists as a lead to check, not a settled fact. Bill numbers, signing dates and effective dates change as bills are amended, and the secondary sources already disagree with each other in places. Before you rely on a specific state, open that state's legislature website and read the enrolled text and its effective-date clause.

How to read a companion-bot notice on screen

The practical effect of these laws is a message that shows up inside an app. Here is what a compliant notice should and should not tell you.

• It should say the character is an AI. In New York that should appear when a session starts and then at three-hour intervals; in California it should be clear and conspicuous.
• It should not be buried. Both laws use the word conspicuous in one form or another. A line in the terms of service does not meet that standard.
• A crisis referral is not a clinical service. The protocols require detection and a pointer to crisis resources. They do not turn a companion app into therapy, and Tennessee's reported law goes in the opposite direction by barring a bot from posing as a licensed mental-health professional.
• A notice is not consent to everything else. It does not change what data the app keeps, how the app uses your conversations or whether it markets to children.

If you are a parent, the minors rules above are the part to read. If you are an adult user, the notice is the whole protection in New York, and you will still want to decide for yourself how long to spend in a conversation. Our own pages on AI companion dark patterns in conversational goodbyes and chatbot minimum age requirements cover the design tricks and age rules that these laws leave alone.

What we could not confirm

• The exact text of California's enrolled SB 243. The legislature's site refused our automated request, so the table relies on a law-firm summary.
• Exact effective dates for the 2027 states. Tech Policy Press gives "2027" only.
• Any court or regulator outcome under either law. We are not aware of one and are not reporting one.

Sources and method

This is an analysis of published legal summaries and reporting, not a test of any product. We compared the statements below against each other and flagged where they disagree.

• Morrison & Foerster, New York and California Enact Landmark AI Companion Laws (client alert, November 2025)
• Tech Policy Press, What 12 State 'Companion Bot' Laws Demand of AI Providers (state list and Colorado rulemaking)
• Epstein Becker Green, Novel AI Laws Target Companion AI and Mental Health (Oregon, Nebraska and Tennessee dates)