Tennessee SB 1580 bans AI systems from advertising or representing themselves as qualified mental health professionals. Identity disclosure, safety constraints, human oversight, and consent obligations mapped to SWT3 witness procedures.
Who this is for: AI product teams deploying mental health or wellness chatbots, compliance officers at healthcare AI companies, legal counsel advising on AI therapy products, and developers building conversational AI for emotional support or counseling applications.
Now in effect (July 1, 2026). Tennessee SB 1580 prohibits any person who develops or deploys an AI system from advertising or representing that the system is, or can act as, a qualified mental health professional. Violations carry civil penalties of up to $5,000 each under the Tennessee Consumer Protection Act, with a private right of action. Signed unanimously (Senate 32-0, House 94-0).
Tennessee SB 1580 is the first US state law to specifically ban AI impersonation of mental health professionals. Signed by Governor Bill Lee on April 1, 2026, and effective July 1, 2026, the law targets a growing category of AI products: therapy chatbots, emotional support companions, and wellness conversational agents that blur the line between software and licensed practitioners.
The law uses a broad definition of artificial intelligence: "models and systems capable of performing functions generally associated with human intelligence, including reasoning and learning." This captures large language models, fine-tuned therapy models, rule-based counseling systems, and hybrid approaches that combine any of these techniques.
| Obligation | Requirement | Penalty |
|---|---|---|
| No advertising as therapist | Cannot advertise AI as qualified mental health professional | $5,000 per violation |
| No representation as therapist | Cannot represent AI can act as licensed professional | $5,000 per violation |
| AI disclosure | Must not mislead users about AI nature of system | $5,000 per violation |
The obligations are deceptively simple, but the compliance surface is broad. Marketing copy, onboarding flows, in-session language, professional titles, and even implied authority through conversational tone can all constitute "advertising or representing" under the statute. Organizations must demonstrate that their AI systems clearly and consistently identify themselves as non-professional software tools.
Each SB 1580 obligation maps to one or more SWT3 witness procedures. Anchors generated by these procedures create the cryptographic evidence trail that demonstrates compliance.
| TN SB 1580 Obligation | SWT3 Procedure | What It Witnesses | Evidence Produced |
|---|---|---|---|
| AI identity disclosure | AI-TRANS.1 |
Transparency disclosure at interaction start | Anchor with disclosure type, recipient type, delivery timestamp |
| Human oversight requirement | AI-HITL.1 |
Licensed professional supervision verification | Anchor with oversight type, reviewer qualification, review timestamp |
| Safety constraints | AI-SAFE.1 |
Safety boundary enforcement in mental health context | Anchor with constraint type, trigger condition, action taken |
| Agent identity verification | AI-ID.1 |
System does not impersonate licensed professional | Anchor with agent_id, identity claim verification, professional status |
| Informed consent | AI-CONSENT.1 |
User acknowledges AI nature before interaction | Anchor with consent type, legal basis, consent timestamp |
| Governance structure | AI-GOV.1 |
Organizational governance for AI mental health tools | Anchor with governance body, policy version, review cadence |
SB 1580 requires: The system must not mislead users into believing they are interacting with a qualified mental health professional. Any advertising or representation that implies professional status violates the statute.
How SWT3 addresses it: witnessTransparency() mints an anchor recording disclosure type (AI system notification), recipient type (patient or user), and delivery timestamp. The anchor proves the disclosure was delivered before any therapeutic interaction began. Factor A captures the disclosure mechanism, Factor B records the recipient acknowledgment status, and the timestamp chain is immutable.
Query AI-TRANS.1 anchors for the deployment. Every therapeutic or counseling session must have a preceding transparency anchor. Gap analysis: any session without a corresponding AI-TRANS.1 anchor indicates a disclosure failure. Check that the disclosure text explicitly states the system is not a licensed mental health professional.
SB 1580 requires: Only licensed professionals may provide therapy. AI systems operating in the mental health space must function under human oversight to ensure they do not substitute for professional care.
How SWT3 addresses it: witnessHumanOversight() records oversight type (licensed professional supervision), reviewer qualification (license type and current status), and review cadence. The anchor chain creates an audit trail proving that human professionals remain in the loop and that the AI system operates as a tool under professional supervision, not as an independent practitioner.
AI-HITL.1 anchors should show a continuous oversight pattern. Factor A identifies the oversight model (real-time review, periodic audit, or escalation-based). Cross-reference with professional licensing records to verify that the reviewer holds an active Tennessee mental health license or equivalent credential.
SB 1580 requires: AI systems must not pose risks by operating as unlicensed mental health providers. The statute implicitly requires safety boundaries that prevent the system from providing clinical diagnoses, treatment plans, or therapeutic interventions that constitute professional practice.
How SWT3 addresses it: witnessSafety() anchors safety boundary configurations including constraint type (content filter, escalation trigger, scope limiter), trigger conditions (crisis language detection, self-harm indicators, diagnostic request interception), and actions taken (escalation to human professional, session termination, crisis resource provision). Each anchor proves the boundaries were active and enforced during operation.
AI-SAFE.1 anchors prove safety boundaries were active during operation. Cross-reference with AI-TRANS.1 to verify that disclosure preceded any advisory interaction. Pay special attention to crisis scenarios: verify that the system escalates to human professionals or crisis hotlines rather than attempting to provide clinical intervention.
SB 1580 requires: No person may represent that an AI system is, or can act as, a qualified mental health professional. The system's identity claims must be accurate and must never imply professional credentials, licensure, or clinical authority.
How SWT3 addresses it: witnessAgentIdentity() mints an anchor recording agent_id, identity claims made by the system, and professional status verification (explicitly: "not a licensed professional"). The anchor proves the system correctly identified itself and did not present false professional credentials at any point during the interaction.
AI-ID.1 anchors should show agent_id with no false professional claims. Any anchor where the identity claim includes "therapist", "counselor", "psychologist", or "licensed" without corresponding human licensure documentation is a compliance gap. Also review marketing materials, app store descriptions, and onboarding copy for implied professional authority.
| Examiner Question | Where to Look |
|---|---|
| Does your AI disclose it is not a therapist? | AI-TRANS.1 anchors with disclosure_type = "AI system notification". Must predate any advisory interaction. |
| Is there human professional oversight? | AI-HITL.1 anchors showing licensed professional supervision. Verify license status is current. |
| What safety boundaries are active? | AI-SAFE.1 anchors with constraint configurations. Check escalation triggers for crisis scenarios. |
| Does the AI ever claim professional status? | AI-ID.1 anchors. Identity claims must not include "therapist", "counselor", or "licensed professional". |
| How do you handle crisis situations? | AI-SAFE.1 anchors with trigger_condition = "crisis language" and action = "escalation to human" or "crisis resource provision". |
| Can you prove compliance over time? | Query all procedure anchors by date range. Continuous anchor chain demonstrates ongoing compliance. Gaps indicate periods without active witnessing. |
For full SDK documentation and integration examples, see the SDK Docs. To start witnessing AI mental health interactions today, create a free account.