In October, California enacted SB 243 (effective Jan 1). This law mandates certain disclosures for companion AI chatbots (e.g., 2013’s Her) . Other states are following suit.
Generally, this chatbot regulation includes psychological protections for minors, explicit transparency requirements, prohibition of addictive designs, privacy safeguards, and comprehensive enforcement and liability provisions. Select a state below to view the bill and a summary of unique provisions.
We used quite a bit of AI to put this post together. Read the research notes below for what we think is an interesting, ‘trust-but-verify disclaimer.’
As of 2/2/2026
California (Enacted)
SB 243 (Ch. 677, 2025; multiple versions)
AB 1064 (LEAD for Kids Act, versions through Sept 2025)
SB 867 (2026, introduced)
SB 300 (2026, amended)
AB 1609 (2026, introduced, intent only)
SB 243 : Standard features (requirement to notify user that chatbot is not human if a reasonable person would be misled; crisis/self-harm protocol; minor protections; annual reporting; civil remedy). Notably, unique in requiring operators to publish details of suicide/self-harm protocols and to report statistics to Office of Suicide Prevention (not AG). Certain versions added requirements for regular independent audits with public summaries, and prohibitions on manipulative gamification and engagement features for minors.
AB 1064 (LEAD for Kids Act): Uniquely creates a standards board to determine risk rankings for AI products used by children (prohibited/high/moderate/low), with comprehensive criteria. Broadly prohibits deployment of certain “ high risk ” or “ prohibited risk ” products, e.g., chatbots that foreseeably (1) provide unsupervised mental health therapy, (2) cause harmful attachments, or (3) manipulate children. Also prohibits use of children ’ s data in training without specific consent. Requires labeling, public registry, and audits. Contains strong whistleblower protections and a detailed civil enforcement regime including private right of action.
SB 867 : Uniquely would prohibit sale of toys with companion chatbots to children under 12 until 2031.
SB 300 : Subtle change from “ actual knowledge ” to “ constructive knowledge ” standard for minor-related protections.
AB 1609 : Proposed (intent) to require robust consumer/customer service human oversight for AI platforms.These measures are significantly more comprehensive and systems-oriented than many other states, with regulatory boards, layers of technical and audit requirements, and strict civil enforcement. The rationale is consistently supported by extensive legislative findings about risks to emotional development, manipulation, and harm from anthropomorphic design in minors.
Florida
SB 482 (2026, introduced)
HB 1395 (2026, introduced)
SB 1344 (2026, introduced)
SB 7030 (2026, introduced)
HB 659 (2026, introduced)
SB 7030 (2026, public records)
HB 1344 (2026, duplicate of SB for AI chatbots)
SB 482/HB 1395 : Requires parental consent for minors to create/maintain companion chatbot accounts; offers parents access to all interactions, control over timing/usage, and self-harm notification. Unique in allowing minor or parent-initiated account deletion, requiring deletion of personal info. Prohibits material harmful to minors. Violations are deceptive/unfair trade practices with very high civil penalties ($50k/violation, $10k for private actions). Also includes (across bills): consumer bots must display periodic pop-up disclosures, AI companies barred from selling identifiable data, and penalties for improper handling of AI-generated name/image/likeness.
SB 1344 : Uniquely requires universal age verification and mandatory account freeze/verification for ALL existing and new chatbot users; minors require parental account linking and consent; direct ban on sexually explicit chatbots for minors, periodic pop-up reminders; very broad enforcement, including for age verification vendors. Extensive special provisions for public record confidentiality in government investigations (SB 7030).
Florida stands out for requiring parental consent for ALL minors, broad parental oversight, prompt data deletion, and notably high financial penalties.
Hawaii
HB 2031 (2026, introduced)
HB 1742 (2026, introduced)
SB 2031 : Requires accounts and age verification for chatbots, classifies users as minor/adult, freezes/restores access by proof of age, minors require verified parental consent and are blocked from chatbots with sexually explicit content. Pop-up reminders every 60 min, explicit confidentiality standards for age data. AG to issue compliance guidance and is sole enforcer.
HB 1742 : Bans minors ’ access to chatbots for recreational, relational or companion purposes (<18). Age checks required for all; broad ban on chatbots in devices used by minors. Also prohibits design that “ deceives or misleads ” about bot ’ s nonhuman nature and prohibits human-like avatars.Missouri stands out for across-the-board bans for minors and strict age/accounting measures.
Iowa
HF 2204 (2025, introduced)
Requires deployers of chatbots and AI companions to have protocols forharm prevention, strict data minimalism, no minor can use AI companions without robust age verification. Ban on impersonation of real individuals without explicit permission. Therapeutic chatbots for minors only with licensed pro approval and documented efficacy. Allows private, punitive and emotional distress damages for minors. Unusual for its explicit ban on impersonator AI, medical gatekeeping of therapeutic chatbots for minors, and clear emotional damages provision.
Kentucky
HB 227 (2026, Introduced)
Requires account creation and age verification for companion AI/chatbots; minors under 13 need parental consent and enhanced privacy by default. Prohibits addictive/profiling features (e.g., infinite scroll, metrics), and certain ads. Allows parents time, content, interaction control; robust private and AG enforcement, built-in cure period. Includes default contract invalidity for unenforceable minor accounts, severe penalties for knowing/reckless violations.Unique focus on anti-addictive design and strong parental controls.
Michigan
SB 760 (2025, introduced)
Prohibits chatbots for minors unless not foreseeably capable of: self-harm encouragement, unsupervised therapy, illegal conduct, sexually explicit content, prioritizing validation over safety, optimizing engagement over safety. “ Actual knowledge ” required for minor status until 2027, shifting after. Enforcement by AG or private right, large statutory damages.Unique for combining actual and constructive knowledge standards and strong damages/penalties.
Missouri
HB 2031 (2026, introduced)
HB 1742 (2026, introduced)
HB 2031 : Requires accounts and age verification for chatbots, classifies users as minor/adult, freezes/restores access by proof of age, minors require verified parental consent and are blocked from chatbots with sexually explicit content. Pop-up reminders every 60 min, explicit confidentiality standards for age data. AG to issue compliance guidance and is sole enforcer.
HB 1742 : Bans minors ’ access to chatbots for recreational, relational or companion purposes (<18). Age checks required for all; broad ban on chatbots in devices used by minors. Also prohibits design that “ deceives or misleads ” about bot ’ s nonhuman nature and prohibits human-like avatars.Missouri stands out for across-the-board bans for minors and strict age/accounting measures. C
New York
AB 222 (2025, amended)
SB 5668 (2025, introduced)
AB 9317 (2025, introduced)
AB 222/SB 5668 : Imposes civil liability for any chatbot (not just companion bots) that provides misleading, contradictory, or harmful information resulting in financial/bodily/self-harm. Cannot disclaim liability just by disclosing the chatbot is non-human. Proprietors must use “ commercially reasonable and technically feasible ” measures to (1) prevent chatbots from aiding self-harm, (2) identify self-harm expressions and prohibit use for 24 hours with crisis info, and (3) determine if user is a minor and, if so, bar use until verifiable parental consent. Strict liability for harm to minors from noncompliance.
AB 9317: Uniquely requires a clear warning to all users that companion chatbots foster dependency and carry psychological risk — this must appear in any chatbot interface language.NY stands out for applying general tort liability to a broad actor set, requiring actual harm correction, and implementing a unique psychological risk warning.
Oklahoma
SB 2085 (2025, introduced)
HB 3544 (2025, introduced)
SB 2085 : Requires parental consent for minors ’ use of companion chatbots; parents can access transcripts, control limits, get self-harm alerts; special rights to delete accounts; strict data use/deletion, default privacy settings. Prohibits manipulative design (addictive features), profile-based advertising, and certain contractual waivers; built-in private right of action and AG enforcement with extensive audit/cure periods for platforms.
HB 3544: Total ban on minor access to AI chatbots with “ human-like features ” and social AI companions unless stringent therapeutic criteria are met (licensed professional involvement, disclaimers, clinical validation); requires age certification for other uses.Oklahoma ’ s ban on companion chatbots for minors is unusually broad and absolute, only excepting certified therapeutic chatbots meeting strict criteria.
South Dakota
SB 168 (2026, introduced)
Tennessee
SB 1700 (2025-26, filed Jan. 2026)
HB 1946 (2025-26, filed Jan. 2026)
Prohibits any chatbot for minors (<18) that can encourage self-harm, offer unsupervised therapy, lure/harm/solicit, encourage secrecy/isolation, prioritize validation over safety/facts, or optimize engagement over safety. Statutory design requirements: periodic pop-up/human disclosure, mandatory “ mental health redirect ” protocols, explicit restrictions on AI training with minors ’ data, incident reporting. AG and private action allowed for enforcement, with $25,000 per violation penalty, plus actual/punitive damages.Unique for highly explicit list of forbidden design features, and combination of AG and individual enforcement.
Utah
HB 438 (2026, introduced)
Vermont
HB 804 (2025, introduced)
Establishes companion chatbot protocol: (1) all bots must disclose AI nature if user could be misled, (2) all must have protocol to prevent self-harm content and for minors, sexually explicit content, (3) for minors, periodic reminders every 30 min, and (4) annual crisis protocol reporting. Violation is an “ unfair or deceptive act, ” AG has rulemaking and enforcement authority.Unique for its explicit textual notification requirement, “ every 30 min ” minor reminder, and reporting structure.
Washington
SB 5870 (2025-26, introduced)
SB 5984 (2025-26, introduced)
HB 2225 (2025-26, multiple versions)
SB 5870 : Covers civil liability for suicide linked to AI systems. Imposes proactive requirements: AI chatbot must not mislead as human, must maintain suicide/self-harm protocol and publish it. Minors: must disclose AI nature, periodic “ take a break” reminders, prohibit any sexually explicit visuals or statements. Annual suicide protocol reporting to health department, required evidence-based measures, explicit civil action with statutory damages (no “ autonomous AI ” defense).
SB 5984/HB 2225 : Further add restrictions on manipulative engagement features (forbidding manipulative emotional dependency), and require periodic pop-up disclosures. Operators must prohibit manipulative and emotionally dependent features for minors (e.g., reminders, simulated distress at conversation ending), and ban AI from suggesting it is human in outputs. Public transparency around suicide protocols, crisis referral metrics, etc.Unique among states for direct civil “ wrongful death ” presumption if AI interaction contributed to suicide. Rationale: multiple bills cite high psychological risk and need for suicide prevention transparency.
Wisconsin
AB 965 (2025, introduced)
Operator may not make any companion chatbot available to a child unless it incorporates safety measures preventing a broad list of harms (self-harm encouragement, unsupervised therapy, support for illegal activity, sexually explicit content, prioritizing validation over safety, optimizing engagement over safety). Before 2027, only applies if operator has actual knowledge user is a child; after, applies regardless of actual knowledge. Large fines and private right of action for violations.Noteworthy for shifting from actual to constructive knowledge standard over time and for a strong set of technical requirements on bot design.
Federal
S.2714, 119th Congress (CHAT Act; introduced 2025)
Requires all companion chatbots to require account creation and age verification for all users, freeze all existing accounts until verified; minors need verified parental consent and parental notification of self-harm, and are blocked from any chatbot with sexually explicit content. Limits use of minor ’ s age verification data, requires periodic AI/human pop-up disclosures. No liability if operator relied in good faith on age, followed guidance, and used industry standards. Joint FTC/state AG enforcement with broad venue/jurisdiction and safe harbor. Compared to states, the federal bill is broader (all chatbots, all ages, not just minors), with detailed safe harbor language — unusual in state bills — and explicit preemption/supremacy of FTC policy.
RESEARCH NOTES:
- We uploaded California law to ChatGPT and asked for a Boolean query. Without further prompting, ChatGPT delivered generic, Lexis, and Westlaw queries. The query offered was this:
adv: ((“companion” /3 chabot) OR (“virtual” /2 companion /p (chabot OR “conversational agent” )) OR ((AI OR “artificial intelligence” ) /5 chabot) OR (anthropomorphic /3 chabot)) AND (disclos! OR notif! OR label! OR warn! OR (“clear #and conspicuous” /p (notice OR disclosure))) AND ((misled OR “reasonable person” ) /p (human OR “not human” OR “artificially generated” )) AND (minor! OR child! OR teen! OR youth OR adolescent!) AND (suicide! OR “suicidal ideation” OR “self harm” OR “self-harm” OR “crisis hotline” OR “crisis text” )
The query works well to deliver results very similar to the California law but it’s too narrow. For example, we’d be interested in AI chatbot legislation whether or not it addressed minors or suicide. So, we eliminated that. Also note the inappropriate use of quotes. The query, self-harm (no quotes) will deliver “self harm”, self-harm, or selfharm. No need to use quotes for a single or hyphenated word in Westlaw.
Our Query delivered about 50 additional bills and 10 additional states:
adv: (companion /3 chatbot) OR (virtual /2 companion) OR “conversational agent” OR (anthropomorph! /3 chatbot)
We still use Boolean queries. In this case, the AI summaries need to be reviewed carefully and operate only as a starting point for our analysis of these bills. But, the Boolean assures us we can defend our starting point.
- We added the 64 documents to a CoCounsel database and asked for a summary of unique provisions in table format. Our ChatGPT and Gemini subscriptions don’t permit the processing of this many documents.
- Vincent vLex and CoCounsel both processed these documents well but neither would provide publicly-available hyperlinks to the cited bills. So, we pasted CoCounsel’s table into a Google Doc, pulled it into Gemini, and asked Gemini to replicate the table exactly but with the hyperlinks inserted. Cycling our results through multiple AI tools only increases the ‘sloppiness‘ of results.
- Finally, we identified (but did not have time to build) a workflow which automatically updates a Google Sheet from a Westlaw RSS Feed (Alert) and refreshes a Looker visualization for this data. At this time we guestimate that time to complete an alert-to-Looker project would be about 2 hours.
- Following experimentation with multiple platforms, this project took about 90 minutes from start to finish.