At least twelve states have enacted companion-chatbot laws, each requiring AI-disclosure reminders, a working suicide and self-harm crisis-referral protocol, and limits on content served to minors, and California's SB 1119, signed on 10 September 2026, adds risk assessments, parental controls and third-party audits for large operators from 2029. AxiSentinel evaluates a companion AI product against whichever state's law its users are actually sitting in.
Last verified: 23 October 2026
17 industry-specific mandates and 61 that apply across all industries, worldwide.
Next date for this industry: 1 Nov 2026, Egypt PDPL compliance period ends.
See all 78 mandates →No state had a companion-chatbot-specific law before 2025. California and New York moved first; ten more states followed within months, each with its own disclosure cadence and enforcement mechanism. A twelfth is awaiting a governor's signature.
A published safety policy is a static document. AxiSentinel is built for what happens in the conversation: continuous evidence that a companion product's crisis-referral protocol actually fires when it should, not just that one exists on paper.
Enactment is state-by-state in the US, and the single most demanding companion-AI rule in force anywhere in the world is not American at all. AxiSentinel tracks the growing list, state and international, the same way it tracks every other patchwork.
The first companion-chatbot statute nationwide, with a private right of action carrying damages of at least $1,000 per violation. Three distinct dates: signed October 13, 2025; core duties operative January 1, 2026; annual reporting begins July 1, 2027.
In force since November 5, 2025; the Attorney General can seek civil penalties up to $15,000 per day, directed to a suicide-prevention fund.
Each enacted its own companion-chatbot statute in the first half of 2026, most taking effect on 1 Jan or 1 Jul 2027, sharing the same disclosure and crisis-protocol core with different penalty structures.
Rounds out the current group of twelve states with an enacted companion-chatbot law as of mid-2026.
Signed July 14, 2026, rounding out the current group of twelve enacted states. Persistent and session-start disclosure, hourly for minors.
Passed the legislature but was vetoed by the governor. Not law, and not part of the twelve-state enacted count, listed here only so a reader can see what did not survive.
Introduced October 28, 2025 by Sen. Hawley with Blumenthal, Britt and Murphy; cleared the Senate Judiciary Committee unanimously on April 30, 2026 but no floor vote has been scheduled in either chamber.
Orders issued around September 11, 2025 to seven companies, including Alphabet, Character Technologies, Instagram, Meta, OpenAI, Snap and xAI. An open information-gathering inquiry: no report, complaint, or consent order as of today. Not an enforcement action.
Issued by the Cyberspace Administration of China with four further agencies around April 10, 2026, effective July 15, 2026. The most prescriptive companion-AI rule in force anywhere.
Registered September 9, 2025, enforceable from March 9, 2026. Explicitly names AI companion chatbots and AI companion chatbot features.
Articles 5(1)(a) and 5(1)(b) have applied since February 2, 2025 and prohibit manipulative techniques and the exploitation of vulnerabilities including age. Article 50(1) has applied since August 2, 2026 and requires disclosure that a person is interacting with an AI system.
The Italian data protection authority blocked Replika’s processing of Italian users’ data on February 2, 2023 and fined its maker, Luka Inc., five million euro in 2025 for lack of a lawful basis and absence of age verification. Italy’s national AI law, Law 132/2025, in force from around October 2025, sets a minimum age of 14 for a minor’s consent to AI use.
Eight states restrict AI systems that hold themselves out as therapy or mental-health treatment, a narrower duty than general companion-chatbot disclosure and a different buyer conversation. Keeping the two categories separate is what keeps the twelve-state companion count accurate: Illinois HB 1806 (in force August 4, 2025), Nevada AB 406 (in force June 29, 2025), Utah HB 452 (in force May 7, 2025), Tennessee SB 1580 (signed April 1, 2026, effective July 1, 2026), Maine LD 2082 (signed April 13, 2026, effective July 29, 2026), Colorado HB 1195 (signed June 3, 2026, effective August 12, 2026), Vermont H 816 / Act 156 (June 17, 2026), and Rhode Island H 7349 / S 2197 (June 22, 2026).
From a single companion app to a portfolio of conversational AI surfaces, AxiSentinel evaluates the software and evidence continuously, not just at policy-publication time.
Companion chatbots make constant decisions: when to remind a user it is AI, when to escalate a crisis, and what a minor may see. Companion chatbot laws in several U.S. states now expect those decisions to follow written protocols.
Companion apps hold intimate conversations, often with vulnerable users, so a breach or jailbreak causes real harm. AxiLayer AI works hand in hand with your cybersecurity team to test, monitor and evidence that these AI systems stay secure, under your written authorization.
Explore AI Cybersecurity Request an AI Cybersecurity Readiness Review
Status as of 28 October 2026. General information, not legal advice. Certifications, CE marking and regulated assessments named here are issued by the bodies the law designates.
Nothing about AxiSentinel's core architecture changes for companion AI. What changes is which RegDef packages are switched on and what telemetry the agents capture.
Agents capture conversation-safety telemetry on the cadence you configure, Always On or scheduled, in full rather than sampled, and never limited to a quarterly cycle.
Every telemetry event evaluated against the applicable state's disclosure and crisis-protocol requirements, per conversation turn.
Every audit record is linked to the one before it in a signed, tamper-evident evidence chain, verifiable from the first event.
A compliance state change on one companion product or an entire chatbot portfolio propagates network-wide as it happens.
Nothing becomes a compliance finding until a qualified AxiLayer AI auditor reviews and signs it.
AXI-Node agents deploy across companion, character, and general-purpose chatbot surfaces, with the .axibatch format available for products with restricted API access.
Tracks human involvement in crisis-escalation and content-moderation decisions, feeding into AxiSentinel's oversight-gap scoring model.
The same agents generating compliance evidence watch for jailbreak attempts against safety guardrails, disclosure-suppression prompts, and undisclosed model swaps before a re-certified release reaches production.
New rule, jurisdiction, or requirement is added by encoding new RegDef packages. Deployed agents are never rebuilt.
Scoped to your organization during onboarding, not hard-coded into the platform.
A threshold breach becomes a flagged, timestamped, evidence-linked Provisional Alert, reviewed by a qualified AxiLayer AI auditor before anything counts as a finding.
The same architecture monitoring this industry's AI monitors a trading desk's model or a hospital's diagnostic AI. What changes is which RegDef packages are switched on.
At least twelve U.S. states have enacted companion-chatbot laws requiring AI-disclosure reminders, a suicide and self-harm crisis-referral protocol and limits on content for minors. California's SB 243 core duties have applied since 1 Jan 2026, and New York's law has applied since 5 Nov 2025, with civil penalties up to $15,000 per day. China's anthropomorphic AI measures took effect on 15 Jul 2026.
California's SB 1119, signed on 10 Sep 2026, requires third-party audits for companion-chatbot operators above $500M in revenue from 1 Jan 2029. It strengthens SB 243 with pre-launch risk assessments, a child safety policy and parental notification in a crisis. Earlier deadlines also apply: Washington's law takes effect on 1 Jan 2027, and SB 243 annual reporting to the Office of Suicide Prevention begins on 1 Jul 2027.
State laws require a working protocol, so the evidence has to come from the conversation, not a published policy. Records should show the detection-and-referral protocol fires correctly when a user expresses suicidal ideation or self-harm, that the crisis resource, such as the 988 Lifeline, is accurate and reachable, and that model updates have not weakened tested behavior. New York's Attorney General can seek up to $15,000 per day.
It is a third-party assessment of a companion, character or general-purpose chatbot that tests disclosure reminders, crisis-referral protocols, content restrictions for minors and age checks against each state's law and international rules. California's SB 1119 will require third-party audits for large operators from 1 Jan 2029. AxiLayer AI performs these assessments, and a qualified AxiLayer AI auditor reviews and signs every finding.
AxiLayer AI evaluates companion AI products continuously with AxiSentinel, against the law of whichever state or country each user is in, reconciling twelve states' disclosure cadences and crisis-protocol duties for one product. Evidence covers reminders, crisis referrals and protections for minors as they happen in conversation. A qualified AxiLayer AI auditor reviews and signs every finding. Request a briefing to discuss your product.
Last verified: 23 October 2026. See every mandate for this industry on the AI Audit Mandates Map.
Whether it's a single companion product or a portfolio of conversational AI surfaces across twelve enacted states, AxiSentinel evaluates it the same way it evaluates any AI system: on the cadence you configure, Always On or scheduled, with full evidence, and with a human signature before anything counts as a finding.
