Companion chatbot laws · what to count and keep · updated

California SB 243: companion chatbots

Since 1 January 2026, an operator of a companion chatbot in California has to show a notice where users could think it is a person, keep and publish a protocol that refers users to crisis services, and, from 1 July 2027, report each year to the Office of Suicide Prevention “The number of times the operator has issued a crisis service provider referral notification” (Cal. Bus. & Prof. Code § 22603(a)(1)) in the year before (read 9 October 2026). The record each duty implies: the notice's text by release, each protocol version with its date, and a count of every referral notification by day or month.

Each duty, in the statute's words

Cal. Bus. & Prof. Code §§ 22601–22605 (SB 243, Chapter 677, Statutes of 2025); in force since 1 January 2026; the annual report from 1 July 2027.

a notice that the chatbot is not human, where a reasonable person could be misled

“an operator shall issue a clear and conspicuous notification indicating that the companion chatbot is artificially generated and not human”

Cal. Bus. & Prof. Code § 22602(a)

Applies from 1 January 2026. To keep: the notice's text and where it shows, each release that changed it.

a protocol against suicidal ideation, suicide or self-harm content, including a notification that refers the user to crisis services

“a notification to the user that refers the user to crisis service providers, including a suicide hotline or crisis text line”

Cal. Bus. & Prof. Code § 22602(b)(1)

Applies from 1 January 2026. To keep: the protocol document, each version with the date it took effect, and a count of each referral notification it showed.

the protocol's details published on the operator's website

“The operator shall publish details on the protocol required by this subdivision on the operator's internet website.”

Cal. Bus. & Prof. Code § 22602(b)(2)

Applies from 1 January 2026. To keep: the address of the published details and the dates each version was up.

for users the operator knows are minors: disclose that they are talking to AI, and by default at least every three hours remind them to take a break and that the chatbot is not human (users known to be minors)

“at least every three hours for continuing companion chatbot interactions that reminds the user to take a break”

Cal. Bus. & Prof. Code § 22602(c)(2)

Applies from 1 January 2026 until 31 December 2026. To keep: the reminder interval for known minors, and the sessions that ran past it without one.

a disclosure, wherever the platform is used, that companion chatbots may not be suitable for some minors

“that companion chatbots may not be suitable for some minors”

Cal. Bus. & Prof. Code § 22604

Applies from 1 January 2026. To keep: where the disclosure shows, on each app, browser or other way in.

evidence-based methods for measuring suicidal ideation

“An operator shall use evidence-based methods for measuring suicidal ideation.”

Cal. Bus. & Prof. Code § 22603(d)

Applies from 1 January 2026. To keep: the method the protocol names and the evidence for it, from your clinical lead.

What changes on 1 January 2027

SB 1119 (Chapter 190, Statutes of 2026) deletes the duties in § 22602(c) for users an operator knows are minors: the AI disclosure and the three-hour break reminder. As a statute of the 2026 session it takes effect on 1 January 2027, and its own child-safety duties apply from 1 July 2027: SB 1119, Adam's Law.

The yearly report

From 1 July 2027, each year, to the Office of Suicide Prevention: the count of referral notifications in the calendar year before, and two sets of protocols, with no user identifiers. The first counts 2026: the SB 243 annual report, what to count and from when.

Enforcement

A person injured in fact may sue for an injunction, the greater of actual damages or $1,000 per violation, and attorney's fees (§ 22605).

The disclosure duty beside California's BOT Act: California chatbot disclosure.

A hosted record of your crisis-referral counts and notice cadence, with the Oregon and Washington public pages kept current for you, is not built

It would take your daily totals (never a conversation or anything about a user), keep them in a hash-chained log with each protocol version, and keep California's yearly draft and the Oregon and Washington pages current.

What exists today: these pages, and the counts and documents you keep yourself. okgate's hash-chained log records the tool calls AI agents make, not your app's events, so it cannot keep this record today.

I want that record

Sources, each read on 9 October 2026