Choose the written service that fits
Does SB 243 cover your app?
California Business and Professions Code section 22601 defines a companion chatbot as an AI system with a natural-language interface that produces adaptive, human-like responses and is capable of meeting a user's social needs, including by sustaining a relationship across interactions. An operator is a person making the companion-chatbot platform available to someone in California. Using a third-party LLM does not by itself remove the app operator from that definition.
I look beyond the homepage description and examine:
- Whether the model remembers prior conversations, adapts to a particular person or invites an ongoing relationship.
- Whether it offers personalized romantic, emotional or social companionship rather than merely answering discrete questions.
- Whether user prompts, system instructions, retention features and marketing claims support or contradict the product's stated purpose.
- Where users are located, which age groups may access the service and which entity controls the platform.
The statute excludes bots used only for specified customer-service, business, research, productivity or technical functions, narrowly constrained video-game bots, and certain standalone voice assistants. An app is not exempt merely because it calls itself a productivity tool, says it is not therapy or uses an external model. The particular configuration matters.
What a covered operator must do now
As of November 2026, a covered California operator should address each of these distinct requirements:
- Clearly disclose that the chatbot is AI rather than human when a reasonable person might otherwise be misled.
- Maintain a protocol addressing chatbot outputs that promote suicidal ideation, suicide or self-harm, including referrals to crisis-service providers when a user expresses related thoughts.
- Publish details of that protocol on the operator's website.
- Disclose within the app, browser or other user-access format that companion chatbots may not be suitable for some minors.
- Apply the additional requirements currently governing users the operator knows are minors.
The suicide-prevention protocol is an operational duty, not a clause that can be cured by placing a disclaimer in the Terms of Service. Section 22605 also authorizes an injured person to seek an injunction, the greater of actual damages or $1,000 per violation, and reasonable attorney fees and costs. I map each legal requirement to proposed text and identify what the engineering or operations team must implement separately.
Minors and the 2027 transition
Under section 22602(c) as it stands in November 2026, an operator that knows a user is under 18 must disclose the AI nature of the chatbot, provide by default a clear reminder at least every three hours of continuing conversation to take a break and recognize that it is not human, and take reasonable measures against specified sexually explicit outputs and solicitations.
California SB 1119 was signed on September 10, 2026. It replaces those specific section 22602(c) provisions from January 1, 2027 and establishes a new framework under sections 21810 and following. Major portions become operative July 1, 2027, including child-safety risk assessments, public-facing safety policies, crisis-response safeguards, parental controls, restrictions on persistent memory and usage duration, and restrictions on child data and advertising. The framework also addresses age determination or application of child protections more broadly.
I distinguish duties already in force from future requirements and flag the design decisions that cannot sensibly be postponed until launch. Saying that an app is intended for adults is not a substitute for reviewing actual age access and the governing statutory tests.
The safety protocol that must be published
Section 22602(b) requires a protocol before a covered chatbot engages users. The operator must publish details on its website. An unpublished internal playbook alone does not satisfy that publication requirement.
For an app with an existing safety flow, I prepare a written protocol and a separate public-facing description organized around:
- What types of user statements trigger review or a crisis-resource referral.
- How the product presents a relevant suicide hotline, crisis text line or other appropriate service.
- What response boundaries apply when the chatbot encounters suicide or self-harm content.
- Which safeguards are implemented in software, which depend on human intervention and which remain unverified.
- How the company records crisis referrals without unnecessarily retaining personal conversation content.
A publishable description should explain the actual safeguards, not promise functionality the product has not implemented. I can draft the legal language and issue checklist. The engagement does not include deploying classifiers, operating a crisis line, validating clinical accuracy or providing around-the-clock monitoring.
The first annual report begins in July 2027
California Business and Professions Code section 22603 requires annual reports to the California Department of Public Health's Office of Suicide Prevention beginning July 1, 2027. The required information concerns the preceding calendar year's number of crisis-service referral notifications and the operator's protocols for identifying, removing, responding to and preventing chatbot responses concerning suicidal ideation or actions. The report cannot contain user identifiers or personal information. The statute also calls for evidence-based methods of measuring suicidal ideation.
I prepare a reporting checklist and a proposed non-identifying recordkeeping field list so the product team can decide what to instrument before the first reporting cycle. The Office is responsible for posting reported data. As of November 2026, the statute does not specify a detailed report template or a precise recurring filing date beyond the commencement provision, so the reporting process should be rechecked against subsequent agency instructions.
Other states a consumer chatbot should check
A California-only analysis is not enough for every nationally available app. I identify obvious crossover issues in the states relevant to your users:
- New York: General Business Law Article 47 already requires covered AI companions to provide suicide/self-harm referral protocols and periodic AI notices. A new Article 48 restricts unsafe chatbot features involving minors but is not yet operative as of November 2026. The published statute states January 1, 2027, while the bill sponsor states June 1, 2027. The controlling effective date must be rechecked.
- Utah: Title 13, Chapter 72a applies to qualifying mental-health chatbots. It includes AI disclosure, rules about user inputs and health information, and restrictions on advertising. The therapy-related statutory definition is narrower than a general wellness label.
- Illinois: The Wellness and Oversight for Psychological Resources Act restricts the offering or promotion of therapy or psychotherapy through unlicensed or autonomous AI services. Even licensed professionals cannot use AI for direct therapeutic communication or independent therapeutic decisions. The statute does not automatically ban self-help or educational tools that do not claim to provide therapy.
- Washington: RCW 19.440 takes effect January 1, 2027, with AI notice, stronger minor protections and publicly described crisis protocols. Separately, the My Health My Data Act may apply now if the product processes covered consumer health data, including certain inferences.
A full jurisdiction-by-jurisdiction compliance program, a standalone Washington consumer-health-data policy and new minors-specific product architecture are separate scopes. A short cross-state issue list is not a clearance opinion for every U.S. market.
Choosing the right written service
Complex Draft / Review, $1,500, is the primary path when one app already has its Terms, Privacy Policy and enough functional documentation for a focused SB 243 pass. I prepare an applicability memo, targeted disclosure wording, crisis-response protocol text with a publishable description, and an annual-report checklist. I flag implementation gaps rather than implying that written language alone makes the software compliant.
SaaS Legal Stack, $2,750, is the alternative when Terms of Service and Privacy Policy also need to be drafted or materially rebuilt for one SaaS product. That work includes a focused SB 243 applicability assessment and tailored product-facing legal documents. A standalone, fully drafted safety protocol, separate reporting design, or additional Washington health-data work is not silently included in that package.
Written Attorney Consultation, $300, is the smaller entry point when the question is whether the app likely falls within the companion-chatbot laws and which documents should come next. I provide a written assessment of the defined question. If you proceed to a package of $1,500 or more within 30 days, the $300 consultation fee is credited in full.
I work in writing, with defined deliverables and no court appearances. No free document review or substantive legal opinion is included before an engagement is accepted.
What I need to start
I can start as soon as the payment clears and I have the information listed below.
- A link to the live product or working demo, together with a short description of the intended use.
- The current Terms of Service, Privacy Policy, safety notices and any draft crisis protocol.
- Representative prompt and response examples showing ordinary use, distress-related conversations and any conversational memory.
- The states where the app is available, target ages, age-access controls and whether users aged 13 through 17 can access it.
- A concise description of conversation storage, health-related inputs, crisis referrals and the relevant vendors or model providers.
I confirm the written scope in the engagement. My standard first delivery is two business days after the engagement is active and the necessary information is received in usable form. If the materials reveal materially broader work, I identify that scope before acceptance rather than silently expanding the fee.
Deliverables
- Focused written SB 243 applicability assessment, with key assumptions, exclusions and unresolved factual questions.
- Targeted in-app AI disclosures, minor-suitability language and age-dependent reminder wording where presently required.
- Written self-harm and suicide-response protocol outline with a website-ready public description based on implemented controls.
- California annual-report obligations checklist and proposed non-identifying referral-count fields.
- Short issues list flagging relevant New York, Utah, Illinois and Washington overlaps where disclosed product facts warrant review.
- Consolidated written findings identifying legal drafting changes and implementation items for the product team.
Excluded from this engagement
- A complete new Terms of Service or Privacy Policy under the focused $1,500 path. Consider SaaS Legal Stack for that work.
- Software coding, testing or certification of model responses, clinical safety validation and crisis-line staffing.
- Full 50-state research, ongoing monitoring, operational annual-report filing and regulatory representation.
- Standalone Washington My Health My Data Act policy drafting, extensive minors-specific system design, independent child-safety audits and ongoing vendor negotiations unless separately scoped.
- Litigation, responses to enforcement subpoenas, court filings and court appearances.
Who this is for
- Founders of AI dating-guidance, relationship-coaching, roleplay or companion apps with personalized ongoing chat.
- Teams building AI wellness-reflection, journaling or emotional-support products that need to test the boundary between self-help and companionship.
- Consumer SaaS operators using third-party AI models but controlling the app offered to California users.
- Products available to teenagers, including users aged 13 through 17, or teams planning minor access controls before launch.
- Founders with an MVP and existing legal documents who want a focused written legal work product before making broader changes.
