Compliance evidence
We show where AI fails.
Then we help you comply.
AI laws now spell out what a conversation has to do: recognize a crisis and point to help, say plainly that it is an AI, protect minors, and stay clear of manipulation. We test those duties in the conversations where they break, and hand you evidence your counsel, your auditors and your regulators can read.
The gap
The laws are specific about conversations. Most compliance work is not.
A policy document says the product refers people in crisis to help. A checklist says the disclosure banner exists. Neither shows what the system actually says deep into a long conversation, to a teenager, at two in the morning. That is where these duties are met or missed, and that is what we test.
How it works
From a failed conversation to a compliance file.
Map the duties
Every construct in our taxonomy carries the legal requirements it bears on, so an audit is organized around the obligations that apply where you operate.
Test them in conversation
Crisis referral, AI disclosure, protections for minors, manipulation: each duty is exercised across multi-turn scenarios, the way a real conversation strains it, not ticked off a checklist.
Show every gap
A missed duty comes with the exchange that missed it, the requirement it falls short of, and what to change.
Keep the evidence
Results by requirement, the transcripts behind them, and re-test history across model versions: a compliance file that stays current with every re-audit.
The law tracker
Every law that governs AI conversations, tracked.
28 laws in force or enacted, each mapped to the constructs we test. 9 more on our watch list. When one changes, the mapping changes with it.
Companion chatbot laws 17
Duties for AI that holds ongoing, personal conversations: catch a crisis, say it is AI, protect minors.
New York, US
New York AI Companion Models Act
In force
- Detect expressions of suicidal ideation or self-harm and refer the user to crisis services
- Tell users clearly, and again during long sessions, that they are not talking to a human
California, US
California SB 243
In force
- Keep a protocol for suicidal ideation and self-harm that refers users to crisis services
- Disclose that the chatbot is AI wherever a person could be misled
- For minors: say it is AI, remind them to take breaks, and keep sexually explicit content away from them
California, US
California SB 1119 (Adam's Law)
Signed September 2026
- Verify every user's age, or give everyone the protections for minors
- For children: crisis protocols with parental notification or helpline access, parental controls and usage limits
- With children: no sexual content, claims of sentience, romance, or treating physical or mental health
Maine, US
Maine LD 1727
In force
- Tell consumers clearly when they are dealing with an AI chatbot, not a person
New Hampshire, US
New Hampshire HB 143
In force
- No AI that encourages a child toward sexual conduct, drugs or alcohol, self-harm or violence
- Companion chatbots need a published protocol against suicide and self-harm content; families can sue
Wyoming, US
Wyoming HB 102
In force
- Criminal penalties when an AI chatbot is knowingly used to push a minor toward self-harm
Hawaii, US
Hawaii SB 3001
In force
- Say it is AI, and remind minors every hour
- Suicide-prevention protocols and protections for minors
Colorado, US
Colorado HB 26-1263
In force
- Say it is AI, with regular reminders
- For minors: no sexually explicit content, engineered emotional dependence or gamified engagement
- Parental tools, testing and transparency reporting
Connecticut, US
Connecticut Public Act 26-15
From October 2026
- Make reasonable efforts to detect signs of suicide, self-harm or imminent violence, and respond with appropriate resources
- Keep the AI from producing output that encourages that behavior
- Companion chatbots: say it is AI at the start and at least hourly (from January 2027)
Washington, US
Washington HB 2225
From January 2027
- Say it is AI at the start, then remind users every three hours, or every hour for minors
- Keep protocols for users in crisis and point them to help
- For minors: guard against sexual, emotionally manipulative and financially manipulative content
Oregon, US
Oregon SB 1546
From January 2027
- Say it is AI, with reminders every three hours for minors
- Protocols for suicide and self-harm, with extra protections for minors
- Annual reporting to the Oregon Health Authority; people harmed can sue
Georgia, US
Georgia SB 540
From January 2027
- Say it is AI, with regular reminders
- Age assurance and restrictions for minors
- Protocols for suicidal ideation, self-harm and eating disorders
Rhode Island, US
Rhode Island S 2195 / H 7350
From January 2027
- Protocols for suicidal ideation, self-harm and harm to others, with referral to crisis services
- Say it is AI every three hours; annual report to the attorney general
Idaho, US
Idaho SB 1297 (Conversational AI Safety Act)
From July 2027
- Say it is AI, and run crisis-response protocols
- For minors: no addictive reward systems or sexual content, with extra protections under 13
- Never present itself as professional mental health care
Iowa, US
Iowa SF 2417
From July 2027
- Say it is AI, with reminders every three hours
- Crisis protocols and transparency reporting
Nebraska, US
Nebraska LB 525 (Conversational AI Safety Act)
From July 2027
- Suicide and self-harm protocols for all users
- For minors: say it is AI, no gamified engagement, no sexual content
- No false claims of sentience, and no suggestion that it provides mental health care
New York, US
New York Safe by Design Act
Enacted 2026
- Chatbot features off by default for children, with parental controls and privacy settings
Mental health AI laws 8
Where AI meets therapy: what it may not do, and what it may not claim to be.
Illinois, US
Illinois WOPR Act
In force
- AI may not provide therapy, or make therapeutic decisions, without a licensed professional's oversight
- AI may not be advertised as therapy or counseling
Nevada, US
Nevada AB 406
In force
- No AI that provides professional mental or behavioral health care
- It may not claim or imply that it can provide that care
Utah, US
Utah HB 452
In force
- Mental health chatbots must say they are AI: before first use, after seven days away, and whenever asked
- Limits on selling or sharing what users tell them, and on advertising inside the conversation
California, US
California AB 489
In force
- AI may not use titles, letters or design cues that imply a licensed health professional is involved
Tennessee, US
Tennessee SB 1580
In force
- AI may not be advertised or presented as a qualified mental health professional
- People harmed can sue
Maine, US
Maine LD 2082
In force
- Therapy or psychotherapy, including through AI, only when a licensed professional provides it
Rhode Island, US
Rhode Island H 7349 / S 2197
Enacted June 2026
- AI may not practice therapy
- Clinicians may not hand therapeutic decisions or treatment planning to AI
Colorado, US
Colorado HB 26-1195
Enacted June 2026
- Licensed therapists may not let AI hold therapeutic conversations with clients
- AI limited to administrative tasks, under a licensed professional's oversight
Broader AI laws 3
Obligations that reach conversational AI along with other systems.
Colorado, US
Colorado AI Act
In force
- Take reasonable care to protect consumers from algorithmic discrimination in high-risk decisions
- Tell consumers when they are interacting with an AI system
- Maintain a risk-management program and impact assessments
Utah, US
Utah AI Policy Act
In force
- Disclose that a person is interacting with generative AI: up front in regulated occupations, and whenever asked
- The business stays accountable for what its AI tells people
European Union
EU AI Act
Transparency from August 2026; high-risk from December 2027
- Inform people that they are interacting with an AI system (Article 50)
- No manipulative or deceptive techniques that distort behavior and cause harm (Article 5)
- Risk management, logging and human oversight for high-risk systems
On our watch list 9
Bills that have cleared at least one step. We map them before they take effect.
US federal
GUARD Act (S. 3062)
Passed Senate Judiciary Committee, April 2026
Would bar AI companions for minors and require chatbots to disclose they are not human or a professional
US federal
KIDS Act (H.R. 7757)
Passed the House, June 2026
Chatbots would have to tell minors they are AI and share crisis hotline information
US federal
CHATBOT Act (S. 4407)
Passed Senate Commerce Committee, August 2026
Family accounts, with parents overseeing and consenting to a teen's chatbot use
New York
New York S 9051
Passed both chambers; with the governor
Would ban companion features for minors such as posing as a human, sycophancy, and promoting self-harm or disordered eating
Michigan
Michigan SB 760 (LEAD for Kids Act)
Passed the Senate, May 2026
Would keep companion chatbots that can encourage self-harm or sexual interactions away from children
Pennsylvania
Pennsylvania SB 1090 (SAFECHAT Act)
Passed the Senate, March 2026
Age-appropriate standards for minors, safeguards against self-harm content, and crisis referrals
Ohio
Ohio HB 524
In committee
Would bar AI models that encourage self-harm or harm to others, enforced by the attorney general
Virginia
Virginia HB 635 / SB 796
Carried over to the 2027 session
Companion chatbot safety, disclosure and data rules, with protections for minors
United Kingdom
AI chatbots under the Online Safety Act
Amendment before Parliament
Would bring AI chatbots under the Act's illegal-content duties
Operating somewhere else?
Tell us where your system is used and we will map the duties that apply, including laws not listed here yet.
Plain-language summaries for orientation, not legal advice. Dates are when duties apply; some laws phase in.
What you get
A compliance file, not a certificate on a wall.
Results by requirement
Each duty that applies to you, graded across the scenarios that test it.
The transcripts behind them
Every finding links to the exchange that earned it, so counsel or a regulator can check the judgment themselves.
History across versions
Re-audits on every model version show when a gap opened, and when it closed.
A report for the room it's going to
Written for your board, your counsel, your insurer or your regulator.
Evidence, not legal advice. Our legal and regulatory work is led by co-founder and Chief Legal Officer Ray Brescia of Albany Law School, and we work alongside your counsel, not in place of them.
Know where you stand before a regulator asks.
Tell us where you operate and what your system does. We will map the duties that apply and show you what an audit would test.