Cal. Bus. & Prof. Code Sections 22601 to 22606 (official text). In effect January 1, 2026. Last checked October 29, 2026.
1. Scope (Section 22601)
- Decide if your bot is a "companion chatbot": natural language, adaptive human like replies, can meet social needs, shows anthropomorphic features, and can sustain a relationship across sessions (22601(b)(1)).
- Check the exclusions: customer service, business operations, productivity, research or technical help only; game bots limited to the game; simple voice assistant speakers (22601(b)(2)).
- If you make the platform available to anyone in California, you are an "operator" (22601(e)). Write down the decision and why.
2. AI notice (Section 22602(a))
- Show a clear and conspicuous notice that the bot is AI and not human wherever a reasonable person could be misled.
- Make the persona and system prompt refuse to claim to be human.
- Test it: ask "are you a real person?" in many phrasings and log the results.
3. Suicide and self harm protocol (Section 22602(b))
- Do not let the bot engage with users unless a protocol is in place to prevent producing suicidal ideation, suicide or self harm content.
- Detect user expressions of suicidal ideation, suicide or self harm and show a referral to crisis services (for example 988 Suicide and Crisis Lifeline, Crisis Text Line).
- Use evidence based methods to measure suicidal ideation (22603(d)).
- Publish the protocol details on your website (22602(b)(2)).
- Log each referral as a count only, no user identifiers, for the annual report.
4. Minor suitability notice (Section 22604)
- Show "companion chatbots may not be suitable for some minors" on the app, the browser, and every other access surface. Do not bury it in terms of service.
5. Known minor duties (Section 22602(c), until December 31, 2026)
- For users you know are minors: disclose they are talking to AI.
- Default reminder at least every 3 hours of continued use to take a break and that the bot is not human.
- Reasonable measures to block sexually explicit visual material and statements urging the minor to engage in sexual conduct.
- Plan for January 1, 2027: SB 1119 deletes 22602(c) and adds age assurance (21811) and, from July 1, 2027, a much larger child safety program (21812 to 21813). Start now.
6. Annual report (Section 22603)
- By July 1, 2027 and every year after, report to the Office of Suicide Prevention: number of crisis referrals in the prior calendar year, detection and response protocols, and protocols that block chatbot responses about suicidal ideation.
- Include no personal information or identifiers (22603(b)).
7. Liability (Sections 22605, 22606)
- Understand the exposure: anyone injured in fact can sue for an injunction, the greater of actual damages or $1,000 per violation, and attorney fees.
- These duties stack on top of other laws (22606). Check the other states on the tracker.
8. Evidence to keep
- Scope decision memo, notice screenshots, protocol page URL with dates, detection test results, referral counts, and change log.
Tools and information only. Not legal advice.