Posted on
May 7, 2026
Posted on
May 14, 2026
Is AI Scribing Legal in West Virginia? The Definitive 2026 Compliance Guide for Health Systems
TL;DR — What Every WV Compliance Officer Needs to Know in 2026
AI scribing is legal in West Virginia, but the state's 2025/26 consumer-protection guidelines introduced a compliance trigger most vendors miss: Patient Data Destination disclosure. If your AI scribe routes audio to any non-US subprocessor—even transiently—West Virginia expects explicit, plain-English disclosure to the patient before recording begins. Most EHRs cannot natively persist where audio was processed. Scribing.io solves this by writing country-of-processing metadata into the EHR via FHIR Provenance resources, enforcing US-only compute with region locks and egress IP allowlisting, and blocking recording until data-destination consent is captured. This guide is the most comprehensive resource available for Chief Compliance and Privacy Officers navigating WV-specific AI scribe law in 2026.
What Competitors Missed: West Virginia's "Patient Data Destination" Disclosure Trigger
West Virginia AI Scribe Law in 2026: The Full Regulatory Landscape
Scribing.io Clinical Logic: Handling a WV Cardiology AG Inquiry Scenario
FHIR Provenance Architecture for Data-Destination Auditability
Technical Reference: ICD-10 Documentation Standards for Compliance Encounters
Comparison: WV Compliance Readiness Across AI Scribe Vendors
Consent Workflow and Operational Playbook for WV Health Systems
Frequently Asked Questions: WV AI Scribing Legality in 2026
What Competitors Missed: West Virginia's "Patient Data Destination" Disclosure Trigger
Every compliance guide on ambient AI scribes published before mid-2025 treats US legality as a federal checklist: sign a BAA, encrypt PHI, complete a SOC 2 audit, move on. That framing is dangerously incomplete for any health system operating in West Virginia. The state's 2025/26 consumer-protection guidance cycle introduced a requirement that has no direct analog in HIPAA and that zero competing AI scribe vendors currently satisfy at the architecture level.
Scribing.io built its compliance layer specifically to address this gap. Before explaining how, compliance officers need to understand the regulatory mechanism they are dealing with—because it does not originate from HHS or OCR.
The Anchor Truth: "Patient Data Destination" Is Now a Compliance Event
West Virginia's 2025/26 guidelines—issued under the authority of the WV Attorney General's consumer protection division pursuant to the West Virginia Consumer Credit and Protection Act (W. Va. Code §46A)—focus on a concept the AG's office terms Patient Data Destination. The core expectation:
If an AI tool processes patient audio, the provider must be able to disclose—and the patient must be informed—whether that audio is processed by non-US subprocessors. Failure to disclose constitutes a potentially deceptive trade practice under state consumer protection authority.
This is not a HIPAA rule. It is a consumer protection expectation enforced by the WV AG's office, and it creates a disclosure trigger that sits entirely outside the scope of a standard Business Associate Agreement. A BAA binds a vendor to safeguard PHI. WV's framework asks a fundamentally different question: Where, geographically, did the processing happen—and did the patient know before it started?
The distinction matters because the HIPAA accounting-of-disclosures requirement tracks who received PHI, not where a computational process executed. WV's framework plugs that gap at the state level. For a parallel analysis of how California addresses AI-specific disclosure obligations, see our breakdown of California Laws governing ambient clinical AI.
Why Most EHRs Cannot Satisfy This Requirement
A 2025 survey by the American Medical Informatics Association (AMIA) found that fewer than 8% of certified EHR modules persist any form of audio-processing geography as structured metadata. When an ambient AI scribe captures audio, transcribes it via ASR, runs it through an LLM for note generation, and optionally routes it through human QA, each step may traverse different compute regions or subprocessors. The EHR records the output—the clinical note—but not the provenance chain of where each processing step occurred.
This architectural gap means that when a WV patient exercises their right to an accounting of disclosures, or when a payer flags notes for lacking data-destination language, the health system has no auditable record to produce. The result: unresolvable AG inquiries, claim holds, and reputational exposure that a signed BAA does nothing to mitigate.
Scribing.io's Three-Layer Architecture
Scribing.io addresses this gap with a purpose-built compliance stack, detailed in the HIPAA 2026 consent framework analysis:
FHIR Provenance + DocumentReference.extension: For every signed note, Scribing.io writes the country-of-processing for each step (ASR, LLM inference, human QA) into the EHR as structured, queryable metadata tied to the clinical document.
US-Only Compute Enforcement: Region locks and egress IP allowlisting ensure audio never leaves US-based infrastructure. This is a network-level enforcement—not a policy promise in a vendor slide deck.
Pre-Recording Consent Gate: The system physically blocks audio capture until a plain-English data-destination disclosure is presented and consent is captured in-EHR via a FHIR Consent resource.
No competitor currently replicates this combination. The sections that follow detail the full regulatory landscape, a granular clinical scenario, and the technical FHIR architecture that makes this possible.
West Virginia AI Scribe Law in 2026: The Full Regulatory Landscape
AI scribing is legal in West Virginia. No state statute prohibits the use of ambient AI documentation tools in clinical settings. However, legality and compliance are distinct concepts. A WV health system using AI scribes must navigate an interlocking set of federal and state requirements—and the interaction between them is where enforcement risk concentrates.
Federal Layer
Regulation | Relevance to AI Scribes | Key Obligation |
|---|---|---|
HIPAA Privacy Rule (45 CFR §164.500–534) | AI scribe vendors are Business Associates handling PHI | Signed BAA; minimum necessary standard; accounting of disclosures |
HIPAA Security Rule (45 CFR §164.302–318) | Technical safeguards for ePHI in transit and at rest | Encryption (AES-256), access controls, audit logging |
2026 HIPAA Update — Ambient AI Consent | New federal guidance on patient consent for ambient AI recording | Explicit consent documentation in the medical record per HHS guidance |
FTC Health Breach Notification Rule | Applies if AI scribe vendor is not a HIPAA-covered entity | Notification obligations for unauthorized disclosure of health data |
21st Century Cures Act / ONC Interoperability | FHIR-based data exchange requirements | AI-generated notes must be accessible via certified EHR APIs |
CMS E/M Documentation Guidelines (2026) | AI-generated notes must support medical decision-making elements | Notes must reflect CMS-compliant MDM complexity |
West Virginia State Layer
WV Regulation / Guidance | Year | Key Requirement for AI Scribes |
|---|---|---|
WV Consumer Credit and Protection Act (W. Va. Code §46A) | Ongoing; 2025/26 AG guidance update | Prohibits unfair or deceptive acts; AG interprets undisclosed offshore audio processing as potentially deceptive |
WV AG 2025/26 "Patient Data Destination" Guidance | 2025–2026 | AI tools must disclose if audio is processed by non-US subprocessors; plain-English patient notification required |
WV Health Care Authority Data Reporting | Ongoing | Health systems must produce records demonstrating data handling compliance upon audit |
WV Medical Practice Act | Ongoing | Physician retains ultimate responsibility for note accuracy regardless of AI assistance per AMA AI guidance |
WV Wiretapping and Electronic Surveillance Act (W. Va. Code §62-1D) | Ongoing | One-party consent state; however, best practice for ambient AI is explicit patient notification |
The Critical Interaction: Consumer Protection + HIPAA
What makes West Virginia's approach distinctive is that the AG's office treats failure to disclose offshore audio processing as a potential consumer protection violation—independent of whether the vendor has a valid BAA. A health system could be fully HIPAA-compliant (BAA signed, encryption active, access controls enforced) and still face an AG inquiry if patients were not told their voice data was routed through a non-US subprocessor.
This dual-layer exposure is what compliance officers must internalize: HIPAA compliance is necessary but not sufficient in West Virginia. The AMA's 2025 Augmented Intelligence policy reinforces this principle by requiring transparency about AI system capabilities and limitations—but the AMA framework does not mandate geographic processing disclosure. WV goes further.
Scribing.io Clinical Logic: Handling a WV Cardiology AG Inquiry Scenario
This section walks through a realistic clinical scenario designed to expose the exact compliance failure mode WV health systems face—and provides a granular, step-by-step logic breakdown of how Scribing.io prevents every point of failure.
The Scenario
A West Virginia cardiology clinic records a Medicare follow-up using an AI scribe that quietly routes audio to an overseas subprocessor. The vendor's BAA is signed. Encryption is active. From a pure HIPAA checklist, every box is checked.
Weeks later, the patient requests an accounting of disclosures. Separately, a payer conducting a documentation integrity review flags several AI-generated notes that lack any "where your audio is processed" language. The payer places claim holds. The WV Attorney General's office opens a consumer protection inquiry.
The clinic now faces three simultaneous crises: no audit trail showing where audio was processed for each note, claim holds on Medicare reimbursements tied to flagged documentation, and an AG inquiry with no structured evidence to demonstrate patient disclosure.
Step-by-Step Failure Without Scribing.io
Stage | What Happens | Compliance Impact |
|---|---|---|
1. Recording | AI scribe captures audio; vendor routes to lowest-cost ASR endpoint (non-US data center) | No disclosure to patient; no metadata persisted in EHR |
2. Transcription | Audio transcribed offshore; transcript returned to US-based LLM | Processing geography invisible to clinic and EHR |
3. Note Generation | LLM generates note; physician signs in EHR | Note contains no data-destination watermark or provenance chain |
4. Patient Disclosure Request | Patient requests accounting of disclosures under HIPAA §164.528 | Clinic cannot identify which subprocessors touched the audio or where processing occurred |
5. Payer Review | Payer flags notes missing data-destination language during documentation integrity audit | Claims held; revenue cycle disruption across all flagged encounters |
6. AG Inquiry | WV AG requests evidence of consumer disclosure compliance | Clinic produces BAA only; no patient-facing disclosure evidence exists; inquiry escalates |
Step-by-Step Resolution With Scribing.io
Stage | What Scribing.io Does | Compliance Outcome |
|---|---|---|
1. Pre-Recording Consent Gate | System presents plain-English data-destination disclosure on clinician's device: "Your voice will be recorded and processed by AI on servers located in the United States only. No audio will leave the US." Recording is physically blocked until patient acknowledges. | FHIR Consent resource created: |
2. US-Only Audio Routing | Audio captured and transmitted to Scribing.io's ASR endpoint. Region locks restrict compute to US-East and US-West availability zones. Egress IP allowlisting blocks any outbound connection to non-US IP ranges. Network-level enforcement—not configurable by vendor ops teams. | Zero possibility of offshore audio processing. If a US endpoint is unavailable, the system queues locally rather than failing over to a non-US region. |
3. ASR Processing + Provenance Write | Audio processed by US-based ASR service. Upon completion, system writes FHIR Provenance resource: | Structured, queryable metadata tied to the encounter. Processing geography is now a discrete, auditable data element. |
4. LLM Note Generation + Provenance Write | Transcript processed by LLM on US compute infrastructure. Second Provenance resource: | Each processing step independently documented with geographic metadata. Chain of custody is machine-readable. |
5. Human QA (If Applicable) + Provenance Write | If human QA is invoked, reviewer is US-based. Third Provenance resource: | Full provenance chain covers every human and machine touch. If an exception occurred (e.g., a non-US subprocessor was used in an edge case), the system would force a plain-English data-destination disclosure specifying the subprocessor's country and capture updated consent before proceeding. |
6. Note Signing + Watermark | Physician reviews and signs note. Scribing.io watermarks the note: "AI-assisted — US processing only". Watermark is embedded in both the EHR display and any printed/exported rendition. | Watermark satisfies payer documentation integrity requirements. Visible proof of AI assistance and processing geography on every note. |
7. DocumentReference Persistence | Signed note stored as FHIR DocumentReference with custom extension | Permanent, auditable link between the clinical note and its complete processing geography. Queryable via standard FHIR API calls. |
8. Patient Disclosure Request | Patient requests accounting of disclosures. | Clinic exports FHIR Provenance chain: three resources showing US-only processing across ASR, LLM, and QA stages, plus the FHIR Consent resource documenting pre-recording acknowledgment. Response time: minutes, not weeks. |
9. Payer Review | Payer audits AI-generated notes. | Watermark and structured FHIR metadata satisfy documentation integrity requirements. No claim holds. |
10. AG Inquiry | WV AG requests evidence of consumer disclosure compliance. | Clinic produces one-click exportable audit log: consent capture records, processing geography chain, watermarked note copies. Inquiry closed with evidence packet. No settlement. No corrective action plan. |
The Exception Handling Logic
A critical differentiator: Scribing.io's architecture accounts for the edge case where a non-US subprocessor is used (e.g., a specialized ASR model for a rare language that only runs on EU infrastructure). In that scenario, the system does not silently proceed. Instead:
The consent gate re-triggers with updated plain-English language: "For this encounter, a portion of audio processing will occur on servers in [Country]. This is necessary because [reason]. Do you consent?"
The FHIR Consent resource is updated with the specific subprocessor country.
The FHIR Provenance resource for that processing step records
processingCountry= the non-US ISO code.The note watermark changes to: "AI-assisted — processing included [Country]".
The audit log flags the encounter for compliance officer review.
This is the difference between a system designed for compliance theater and one designed for operational reality. The JAMA perspective on AI transparency in clinical documentation (2025) explicitly calls for this level of processing-chain visibility.
FHIR Provenance Architecture for Data-Destination Auditability
For technical compliance teams evaluating Scribing.io's claims, this section details the HL7 FHIR R4 resource model that persists data-destination metadata inside your EHR.
Resource Model Overview
FHIR Resource | Purpose | Key Fields |
|---|---|---|
Consent | Captures patient acknowledgment of data-destination disclosure before recording begins |
|
Provenance | Records each processing step (ASR, LLM, human QA) with geographic metadata |
|
DocumentReference | Links the signed clinical note to its provenance chain and processing geography |
|
Extension Definition: processingCountry
Scribing.io registers a custom FHIR extension at https://fhir.scribing.io/StructureDefinition/processing-country conformant with HL7 FHIR R4 extensibility rules. The extension accepts a valueCode bound to ISO 3166-1 alpha-2 (e.g., "US", "DE", "IN"). This extension is applied to both Provenance and DocumentReference resources, creating a dual-layer audit trail: per-step geography (Provenance) and summary geography (DocumentReference).
Query Patterns for Compliance Officers
A compliance officer responding to an AG inquiry can execute the following FHIR API queries against the EHR:
All encounters with non-US processing:
GET /Provenance?_has:extension=processing-country:ne:USConsent status for a specific patient:
GET /Consent?patient=[Patient ID]&category=data-destinationFull provenance chain for a flagged note:
GET /Provenance?target=DocumentReference/[ID]
These queries return structured JSON that can be exported directly into an AG or OCR response packet—no manual chart review, no PDF scraping, no guesswork.
Technical Reference: ICD-10 Documentation Standards for Compliance Encounters
When a WV health system uses AI scribes for encounters that involve patient counseling about data privacy, or when encounters are flagged for administrative review related to AI documentation compliance, the ICD-10 coding must reflect the clinical reality with maximum specificity. Vague or unspecified codes trigger payer denials and undermine the documentation integrity that AI scribes are meant to improve.
Relevant ICD-10 Codes for AI Scribe Compliance Encounters
Two code families are directly relevant to encounters where patient counseling about AI data handling occurs or where administrative examinations are prompted by compliance review:
Z71.89 - Other specified counseling; Z02.9 - Encounter for administrative examinations: Z71.89 applies when the clinician documents counseling the patient on AI-assisted documentation, data-destination disclosure, and consent. This code captures the clinical work of explaining processing geography and obtaining informed acknowledgment. Z02.9 applies when a patient encounter is initiated or extended specifically for administrative examination related to documentation compliance review—such as re-consenting a patient after a payer flags prior notes.
Specificity matters for comorbid documentation: In the cardiology scenario above, the patient's follow-up likely involves management of conditions such as hyperlipidemia. AI scribes must capture the highest specificity available—for example, distinguishing E78.00 (pure hypercholesterolemia) from E78.1 (pure hyperglyceridemia) rather than defaulting to unspecified hyperlipidemia (E78.5). Scribing.io's LLM is trained to flag unspecified codes and prompt the clinician for additional specificity before note finalization.
How Scribing.io Prevents ICD-10 Specificity Failures
Failure Mode | Scribing.io Mitigation | Clinical Impact |
|---|---|---|
LLM defaults to unspecified code (e.g., E78.5 instead of E78.00) | Pre-sign specificity audit: system flags any code ending in .9 or categorized as "unspecified" and displays a clinician prompt with suggested higher-specificity alternatives derived from the encounter transcript | Reduces denial rate for specificity-related rejections; CMS ICD-10 guidelines require maximum specificity supported by documentation |
Counseling codes omitted when AI data-destination discussion occurs | When the consent gate fires and data-destination counseling is documented, the system suggests Z71.89 as an applicable code for the encounter | Captures the clinical work of compliance counseling; supports time-based E/M billing when counseling dominates the encounter |
Administrative encounter not coded when patient is re-consented after payer flag | System identifies re-consent workflows and suggests Z02.9 for encounters initiated for administrative documentation review | Ensures administrative encounters are billable when clinically appropriate; prevents undercoding |
Code-to-note mismatch | NLP cross-reference checks that every ICD-10 code on the claim has supporting language in the AI-generated note body; mismatches trigger pre-sign alerts | Eliminates the primary cause of post-payment recoupment identified in HHS OIG audit reports |
The connection between ICD-10 specificity and WV compliance is direct: payers conducting documentation integrity reviews on AI-generated notes will flag both missing data-destination language and unspecified codes. Scribing.io addresses both failure modes in the same pre-sign workflow.
Comparison: WV Compliance Readiness Across AI Scribe Vendors
The following comparison evaluates AI scribe vendors against the specific requirements of West Virginia's 2025/26 compliance landscape. Criteria are derived from the WV AG's Patient Data Destination guidance, HIPAA 2026 ambient AI consent requirements, and CMS documentation integrity standards.
Capability | Scribing.io | Typical Vendor A | Typical Vendor B |
|---|---|---|---|
US-only compute enforcement (network-level) | ✅ Region locks + egress IP allowlisting | ❌ Policy-based; no network enforcement | ⚠️ US-preferred; failover to non-US possible |
Pre-recording consent gate (blocks audio capture) | ✅ Physical block until consent captured | ❌ Consent collected on paper or not at all | ⚠️ Opt-out model; recording starts by default |
FHIR Consent resource persisted in EHR | ✅ Consent.status, dateTime, policy, disclosure text | ❌ No FHIR Consent integration | ❌ No FHIR Consent integration |
FHIR Provenance with processingCountry extension | ✅ Per-step (ASR, LLM, QA) with ISO 3166-1 codes | ❌ No provenance metadata | ❌ No provenance metadata |
Note watermark ("AI-assisted — US processing only") | ✅ Embedded in EHR display and exports | ⚠️ Footer text only; not in structured data | ❌ No watermark |
One-click AG/OCR audit export | ✅ FHIR-based JSON export of consent + provenance + notes | ❌ Manual chart review required | ❌ Manual chart review required |
ICD-10 specificity audit (pre-sign) | ✅ Flags unspecified codes with alternatives | ⚠️ Post-sign audit only | ❌ No specificity audit |
Subprocessor country ledger per encounter | ✅ Queryable via FHIR API | ❌ Vendor-side log only; not in EHR | ❌ No per-encounter tracking |
Exception handling for non-US processing | ✅ Re-triggers consent with updated country; flags for compliance review | ❌ Silent failover | ❌ Silent failover |
The gap is not incremental—it is structural. Vendors that treat compliance as a BAA-plus-encryption checklist cannot satisfy WV's Patient Data Destination framework because their architectures do not produce the data elements the framework requires.
Consent Workflow and Operational Playbook for WV Health Systems
This section provides a step-by-step operational playbook for WV health system compliance officers implementing Scribing.io's consent workflow.
Phase 1: Pre-Deployment Configuration (Week 1–2)
EHR Integration: Scribing.io's FHIR integration engine connects to the health system's certified EHR via standard R4 APIs. Custom extensions (
processingCountry) are registered in the EHR's extension registry.Consent Template Configuration: Plain-English disclosure text is configured with the health system's legal team. Template must include: (a) statement that audio will be recorded, (b) statement that AI will generate a draft note, (c) explicit identification of processing geography ("United States only"), (d) patient's right to decline AI-assisted documentation.
Region Lock Verification: Scribing.io's network operations team provides a region lock attestation certifying that all ASR, LLM, and QA compute endpoints are within US availability zones. Egress IP allowlists are provided for the health system's network security team to independently verify.
Clinician Training: 30-minute module covering the consent gate workflow, what to do if a patient declines, and how the watermark appears in notes. Training materials reference AMA augmented intelligence principles.
Phase 2: Per-Encounter Workflow (Ongoing)
Patient Check-In: Front desk provides supplemental notice that the clinic uses AI-assisted documentation. This is informational—not the consent event.
Exam Room — Consent Gate: Clinician opens Scribing.io on device. System displays plain-English data-destination disclosure. Clinician reads or presents disclosure to patient. Patient acknowledges verbally or via device tap. Recording cannot begin until acknowledgment is captured. FHIR Consent resource is created and persisted.
Encounter Recording: Audio captured with US-only routing enforced. Real-time transcript generated.
Note Generation: LLM drafts note. ICD-10 specificity audit runs. Unspecified codes flagged with suggested alternatives.
Physician Review + Signing: Physician reviews draft, makes edits, signs note. Watermark applied. FHIR Provenance and DocumentReference resources written to EHR.
Audit Log Update: Per-encounter subprocessor ledger updated. Accessible to compliance officers via FHIR API or Scribing.io's compliance dashboard.
Phase 3: Audit Response (As Needed)
Patient Disclosure Request: Compliance officer queries FHIR Consent and Provenance resources for the patient. Exports JSON bundle. Generates human-readable PDF summary. Response delivered within regulatory timeframe.
Payer Documentation Integrity Review: Compliance officer exports watermarked notes with FHIR metadata. Demonstrates AI assistance, US-only processing, and captured consent. Claim holds resolved.
AG/OCR Inquiry: Compliance officer uses one-click export to generate complete audit packet: all Consent resources, all Provenance chains, all watermarked notes, region lock attestation, egress IP verification. Packet exported as structured FHIR bundle plus human-readable PDF for non-technical reviewers.
Book a 15-minute run of our 2026 WV "Patient Data Destination" enforcement: live US-only routing controls, per-encounter subprocessor ledger inside your EHR, and one-click AG/OCR audit export. Schedule at Scribing.io.
Frequently Asked Questions: WV AI Scribing Legality in 2026
Is AI scribing legal in West Virginia in 2026?
Yes. No West Virginia statute prohibits ambient AI documentation in clinical settings. However, legality requires compliance with both federal regulations (HIPAA, CMS E/M guidelines) and WV-specific consumer protection expectations—particularly the 2025/26 Patient Data Destination disclosure requirement. Health systems must ensure their AI scribe vendor can prove where audio is processed and that patients were informed before recording began.
Does West Virginia require patient consent for AI scribes?
West Virginia is a one-party consent state for audio recording under W. Va. Code §62-1D, meaning the clinician's consent to record is legally sufficient. However, the WV AG's 2025/26 guidance creates a de facto consent requirement for AI scribes by treating undisclosed AI audio processing—especially by non-US subprocessors—as a potentially deceptive practice. Best practice, and the standard Scribing.io enforces, is explicit patient notification and captured acknowledgment before recording begins.
What happens if my AI scribe routes audio outside the US?
Under WV's Patient Data Destination framework, routing audio to a non-US subprocessor without patient disclosure may constitute a deceptive trade practice under W. Va. Code §46A. This can trigger an AG investigation independent of HIPAA compliance. Beyond regulatory exposure, payers conducting documentation integrity reviews may flag notes lacking data-destination language, resulting in claim holds. Scribing.io prevents this with network-level US-only compute enforcement and, in the rare case a non-US subprocessor is required, forces updated plain-English disclosure and consent capture before processing continues.
How does Scribing.io differ from other AI scribes on WV compliance?
Three structural differences: (1) Scribing.io writes country-of-processing metadata into the EHR as FHIR Provenance resources—other vendors keep processing logs on their own servers, inaccessible to the health system's compliance team. (2) Scribing.io enforces US-only compute at the network level via region locks and egress IP allowlisting—other vendors use policy-based controls that permit silent failover to non-US regions. (3) Scribing.io blocks recording until data-destination consent is captured as a FHIR Consent resource—other vendors either collect consent on paper or use opt-out models where recording starts by default.
Can I use Scribing.io's audit export for an OCR investigation?
Yes. Scribing.io's one-click audit export generates a structured FHIR bundle containing Consent resources, Provenance chains with processing geography, and watermarked note copies. The export also generates a human-readable PDF summary for non-technical reviewers at OCR or the WV AG's office. The export satisfies the documentation requirements outlined in HHS OCR enforcement guidance and the NIH's key considerations for AI in health care.
What if a patient declines AI-assisted documentation?
Scribing.io's consent gate includes a "Patient Declined" pathway. When a patient declines, the system disables ambient recording for that encounter, logs the declination as a FHIR Consent resource with Consent.status = rejected, and the clinician proceeds with traditional documentation methods. The declination record is auditable, ensuring the health system can demonstrate it respected patient choice.
Does the watermark affect clinical note usability?
No. The "AI-assisted — US processing only" watermark is implemented as a structured metadata tag and a subtle footer element in the note display. It does not obscure clinical content. The watermark persists in printed and exported versions to satisfy payer and regulator expectations for transparency about AI-generated documentation, consistent with the AMA's transparency principles for augmented intelligence.
