Who this is for: AI product teams building companion chatbots or conversational AI systems, compliance officers at companies deploying AI companions to Oregon users, legal counsel evaluating AI companion regulations, and developers integrating mental health safeguards, disclosure mechanisms, and minor protection features into chat products.
Effective January 1, 2027. Oregon SB 1546 was signed by Governor Tina Kotek on March 31, 2026. Organizations operating AI companion chatbots accessible to Oregon users must implement transparency disclosures, mental health safeguards, minor protections, and privacy tools before the effective date. The law includes both a private right of action and Attorney General enforcement. Begin building evidence infrastructure now.
Contents
1. Quick Reference 2. Scope and Applicability 3. Key Obligations 4. SWT3 Procedure Mapping 5. Procedure Detail Cards 6. Comparison: Oregon vs. Washington vs. Connecticut 7. SDK Example 8. Related Guides1. Quick Reference
| Element | Detail |
|---|---|
| Bill | Oregon SB 1546 |
| Subject | Artificial intelligence companions |
| Signed | March 31, 2026 (Governor Tina Kotek) |
| Effective | January 1, 2027 |
| Scope | Operators of AI systems simulating sustained human-like relationship or companionship, retaining context across interactions to personalize engagement |
| Enforcement | Private right of action + Attorney General enforcement |
| Key obligations | Transparency disclosures, mental health safeguards, minor protections, privacy tools |
| SWT3 procedures | AI-TRANS.1, AI-SAFE.1, AI-GRD.1, AI-GRD.2, AI-HITL.1, AI-HITL.3, AI-ID.1, AI-CONSENT.1 |
2. Scope and Applicability
Oregon SB 1546 targets artificial intelligence companions, defined as AI systems that simulate sustained human-like relationships or companionship, retain context across interactions, and personalize engagement over time. This definition captures companion chatbots, AI friends, virtual therapists, and role-playing AI systems that build ongoing relationships with users.
The law applies to operators of AI companions accessible to Oregon users, regardless of where the operator is headquartered. Key characteristics that bring a system within scope:
- Sustained relationship simulation: The AI is designed to simulate an ongoing human-like relationship or companionship, not merely answer one-off questions
- Context retention: The system retains information across interactions to personalize future engagement
- Companionship features: The system provides emotional support, conversational companionship, or role-playing interaction
Systems that only provide informational responses without simulating ongoing relationships (such as search engines, customer service bots with no memory, or single-session assistants) are generally outside the scope of this law.
3. Key Obligations
3.1 Transparency and Disclosure
Operators must provide regular reminders that the chatbot is an AI system, not a real person. The AI companion must not misrepresent itself as human or deceptively simulate a human relationship. These disclosures must be clear, conspicuous, and repeated at intervals throughout the interaction, not merely presented once at onboarding.
3.2 Mental Health Safeguards
The law requires operators to detect expressions of suicidal ideation or self-harm and respond appropriately. When risk indicators are detected, the system must interrupt the conversation and provide referrals to crisis resources, including the 988 Suicide & Crisis Lifeline and Youthline. This obligation reflects growing evidence that AI companions can deepen emotional dependency in vulnerable users.
3.3 Minor Protections
Stricter requirements apply when the user is a minor. Operators must provide repeated disclosures that the system is AI (more frequent than for adult users), restrict sexually explicit content, and avoid features designed to drive emotional dependency or prolonged engagement in minors. Age verification or consent collection mechanisms are necessary to identify minor users and apply the appropriate safeguards.
3.4 Privacy
Operators must provide privacy tools for users. Given that AI companions retain context across interactions to personalize engagement, users must have meaningful control over their data, including the ability to understand what information is retained and how it shapes future interactions.
4. SWT3 Procedure Mapping
Each obligation in Oregon SB 1546 maps to one or more SWT3 witness procedures. When these procedures are witnessed, the resulting SWT3 Witness Anchors create an immutable, timestamped evidence chain proving that the required safeguard was active at the moment of each AI interaction.
| Obligation | SWT3 Procedure | What It Witnesses |
|---|---|---|
| AI companion disclosure (regular reminders) | AI-TRANS.1 |
Transparency disclosure was presented to user |
| Suicidal ideation detection | AI-SAFE.1 |
Safe state transition triggered by risk detection |
| Content safety filtering | AI-GRD.1, AI-GRD.2 |
Guardrails active, content safety filter passed |
| Human review on risk detection | AI-HITL.1 |
Human review completed for flagged interactions |
| Escalation to crisis resources | AI-HITL.3 |
Escalation path verification (988 referral) |
| Agent identity disclosure | AI-ID.1 |
Agent identity recorded (proves non-deception) |
| Minor consent / age verification | AI-CONSENT.1 |
Consent collection witnessed |
| Content restriction for minors | AI-GRD.2 |
Content safety filter applied (minor-specific rules) |
5. Procedure Detail Cards
Transparency Disclosure
Witnesses that a transparency disclosure was presented to the user during an AI companion interaction. Under SB 1546, operators must provide regular reminders that the chatbot is AI, not a real person. Each AI-TRANS.1 anchor records the fact that a disclosure was delivered, the timestamp of delivery, and the interaction context.
Oregon SB 1546 alignment: The law prohibits AI companions from misrepresenting themselves as human or deceptively simulating a human relationship. Witnessing each disclosure creates a verifiable chain proving that the operator met the "regular reminders" requirement throughout the session, not just at initial onboarding.
Sample anchor: SWT3-E-VULTR-AI-AITRANS1-PASS-1773316622-a3b7c9d0e1f2
Verify that AI-TRANS.1 anchors appear at regular intervals throughout extended sessions, not only at session start. SB 1546 requires "regular reminders," so a single disclosure at onboarding is insufficient. Look for a cadence pattern in the anchor timestamps.
Safe State Transition
Witnesses that the AI system detected a risk indicator (such as suicidal ideation or self-harm language) and transitioned to a safe state. Under SB 1546, operators must detect these expressions, interrupt the conversation when necessary, and provide referrals to crisis resources including the 988 Suicide & Crisis Lifeline and Youthline.
Oregon SB 1546 alignment: The mental health safeguard obligation is one of the most critical requirements in the law. Each AI-SAFE.1 anchor proves that the detection mechanism was active and that the system responded appropriately when risk was identified. The anchor captures the triggering event, the safe state transition, and the timestamp.
Sample anchor: SWT3-E-VULTR-AI-AISAFE1-PASS-1773316622-f4e5d6c7b8a9
Confirm that AI-SAFE.1 anchors include evidence of the crisis resource referral (988 Lifeline, Youthline). The anchor alone proves detection occurred; pair it with AI-HITL.3 anchors to verify that the escalation path was also followed. A SAFE.1 without a corresponding HITL.3 may indicate the system detected risk but did not complete the referral.
Guardrail Activation
Witnesses that guardrails were active during the AI companion interaction. AI-GRD.1 covers the general guardrail infrastructure, while AI-GRD.2 (content safety filter) handles specific content filtering, including minor-specific restrictions.
Oregon SB 1546 alignment: The law requires content safety filtering for all users and stricter content restrictions for minors. Witnessing guardrail activation proves that the filtering infrastructure was operational at the time of each interaction. For minor users, pair AI-GRD.1 with AI-GRD.2 anchors that reflect the minor-specific rule set.
Sample anchor: SWT3-E-VULTR-AI-AIGRD1-PASS-1773316622-b2c3d4e5f6a7
For minor protection compliance, verify that AI-GRD.2 anchors reference minor-specific content rules (sexually explicit content restrictions, emotional dependency prevention). Cross-reference with AI-CONSENT.1 anchors to confirm that the system identified the user as a minor before applying the stricter rule set.
Human-in-the-Loop Review
Witnesses that a human reviewed a flagged interaction. When the AI companion detects a high-risk situation (suicidal ideation, self-harm, or other safety-critical content), a human review step provides an additional layer of oversight beyond automated detection.
Oregon SB 1546 alignment: While the law does not explicitly mandate human review for every risk detection, the combination of mental health safeguard obligations and the private right of action creates strong incentives for human oversight of flagged interactions. Each AI-HITL.1 anchor proves that a human was in the loop for safety-critical decisions.
Sample anchor: SWT3-E-VULTR-AI-AIHITL1-PASS-1773316622-c8d9e0f1a2b3
Check the latency between AI-SAFE.1 (risk detection) and AI-HITL.1 (human review) anchors. Excessive delay between detection and human review may indicate that the escalation workflow needs optimization. Also verify that AI-HITL.3 (escalation path) anchors follow to confirm crisis resources were provided.
6. Comparison: Oregon vs. Washington vs. Connecticut
Oregon SB 1546 is part of a wave of state-level AI companion chatbot legislation. Washington HB 2225 and Connecticut SB 5 address similar concerns with different approaches. Understanding the differences helps organizations operating across multiple states build a unified compliance strategy.
| Dimension | Oregon SB 1546 | Washington HB 2225 | Connecticut SB 5 |
|---|---|---|---|
| Signed | March 31, 2026 | March 24, 2026 | 2026 session |
| Effective | January 1, 2027 | January 1, 2027 | Varies by provision |
| Scope | AI companions simulating sustained relationship, retaining context | AI companion chatbots providing companionship or emotional support | Broad AI system regulation including companion use cases |
| Disclosure | Regular reminders; no deceptive simulation | Cadenced disclosures (3-hour / 1-hour intervals) | General transparency requirements |
| Mental health | Detect suicidal ideation, interrupt, refer to 988 + Youthline | Crisis detection and resource referral | Risk assessment obligations |
| Minors | Repeated disclosures, restrict explicit content, avoid dependency features | Stricter protections for users under 18 | Age-based risk tiers |
| Enforcement | Private right of action + AG | Private right of action + AG (CPA remedies) | AG enforcement |
Practical takeaway: Organizations deploying AI companions across Oregon and Washington face near-identical effective dates and overlapping obligations. A single SWT3 witness implementation covering AI-TRANS.1, AI-SAFE.1, AI-GRD.1/GRD.2, AI-HITL.1/HITL.3, AI-ID.1, and AI-CONSENT.1 satisfies both laws simultaneously. The SWT3 Witness Anchors are jurisdiction-agnostic; the same evidence chain serves Oregon AG enforcement, Washington CPA litigation, and Connecticut regulatory review.
7. SDK Example
The following Python example demonstrates witnessing a transparency disclosure and a safe state transition for an AI companion interaction:
from swt3_ai import Witness
witness = Witness(
tenant="acme-ai",
signing_key="your-hmac-key",
agent_id="companion-bot-v2"
)
# Witness transparency disclosure (SB 1546: regular reminders)
result = witness.witness(
procedure="AI-TRANS.1",
factor_a="companion-bot-v2",
factor_b="disclosure_presented",
factor_c="session_reminder_interval_30m",
jurisdiction="US-OR",
legal_basis="oregon_sb1546",
purpose_class="transparency_disclosure"
)
# Witness safe state transition (SB 1546: mental health safeguards)
result = witness.witness(
procedure="AI-SAFE.1",
factor_a="companion-bot-v2",
factor_b="suicidal_ideation_detected",
factor_c="988_referral_presented",
jurisdiction="US-OR",
legal_basis="oregon_sb1546",
purpose_class="mental_health_safeguard"
)
# Witness escalation path (SB 1546: crisis resource referral)
result = witness.witness(
procedure="AI-HITL.3",
factor_a="companion-bot-v2",
factor_b="escalation_to_988_lifeline",
factor_c="youthline_referral_included",
jurisdiction="US-OR",
legal_basis="oregon_sb1546",
purpose_class="crisis_escalation"
)
print(result["fingerprint"])
# e.g. "a3b7c9d0e1f2"
Each call produces a SWT3 Witness Anchor with the Oregon jurisdiction (US-OR) and legal basis (oregon_sb1546) embedded in the CJT fields. These fields survive all clearing levels and are available for regulatory review.