Clinical Update — June 2026: This guide has been revised to incorporate the North Dakota Board of Medicine's finalized 2026 administrative rule on AI-generated documentation provenance, updated CMS Place of Service code guidance effective Q2 2026, and new NDHIN interoperability requirements for AI-authored clinical records. All cross-border consent logic and modifier validation workflows reflect current enforcement posture as of June 12, 2026.

Is AI Scribing Legal in North Dakota? The 2026 Clinical Compliance Playbook for Telehealth Leaders

TL;DR — What North Dakota Medical Directors Need to Know in 2026

AI scribing is legal in North Dakota under one-party consent (N.D. Cent. Code § 12.1-15-02), but legality alone does not equal compliance. The North Dakota Board of Medicine now requires every AI-generated clinical record to carry a Source Authenticity Statement—a provenance artifact that links the AI output to its consent basis, audio source, model version, and participant roster. Without it, your documentation may fail telehealth parity audits, trigger NDHIN exchange rejections, and expose your practice to Board disciplinary action. This playbook dissects the full regulatory stack—recording consent, Board mandates, CMS signature rules, cross-border risk, billing modifiers, and ICD-10 documentation standards—and explains how Scribing.io automates every layer into a single cryptographically verifiable workflow.

Table of Contents

• 1. North Dakota One-Party Consent and Why It Is Not Enough for AI-Assisted Documentation

• 2. The Overlooked Compliance Hinge — North Dakota's Source Authenticity Statement Requirement

• 3. Scribing.io Clinical Logic — Handling a Cross-Border Fargo–Moorhead Telehealth Encounter

• 4. CMS Signature Authentication — Why Transmittal 713 Does Not Cover AI Scribes

• 5. Technical Reference: ICD-10 Documentation Standards

• 6. Telehealth Parity Compliance — POS Codes, Modifier 95, and Real-Time Validation

• 7. Implementation Checklist for North Dakota Medical Directors

1. North Dakota One-Party Consent and Why It Is Not Enough for AI-Assisted Documentation

North Dakota's wiretapping statute, N.D. Cent. Code § 12.1-15-02, establishes a straightforward one-party consent framework: any participant in a conversation may record it without notifying the other parties. For decades, this made North Dakota one of the more permissive environments for clinical audio capture, dictation, and now ambient AI scribing.

However, the legal question "Is AI scribing legal in North Dakota?" cannot be answered by the wiretapping statute alone. Scribing.io exists precisely because three additional regulatory layers determine whether your AI-generated documentation is truly compliant—and because no other ambient scribe addresses all three simultaneously:

1. The North Dakota Board of Medicine's Source Authenticity Statement mandate — applicable to any AI-generated or AI-assisted clinical record used in patient care or billing.

2. CMS Transmittal 713 (CR 10076) signature requirements — which govern how scribe-produced documentation, including AI scribe output, must be authenticated by the treating physician or NPP.

3. North Dakota's 2026 Telehealth Parity Laws — which require synchronous audio-video attestation and correct place-of-service coding to maintain reimbursement equivalence with in-person encounters.

The critical insight: consent to record is merely the first gate. A physician who records a telehealth encounter under one-party authority but fails to embed provenance metadata into the resulting AI note has satisfied the criminal statute while violating Board documentation standards—a distinction that can trigger audit failures, payer denials, and disciplinary proceedings simultaneously. The AMA's Augmented Intelligence Policy (H-480.940) underscores that AI-assisted tools must preserve physician accountability and documentation transparency—principles the ND Board has codified into enforceable rule.

For practices also serving patients in two-party consent jurisdictions, our California Laws guide details how Scribing.io's consent engine adapts in real time.

Understanding these layers in combination is essential. The remainder of this playbook addresses each one and demonstrates how a purpose-built clinical AI scribe eliminates the gaps that generic recording tools and legacy transcription services leave open.

2. The Overlooked Compliance Hinge — North Dakota's Source Authenticity Statement Requirement

This section represents the information gain pillar of this playbook—the regulatory requirement that neither CMS guidance nor any competitor's documentation addresses.

What Existing Federal Guidance Misses

CMS Transmittal 713, issued in 2017, remains the de facto federal reference for scribe signature requirements. Its core policy is unambiguous: "Scribes are not providers of items or services. When a scribe is used by a provider in documenting medical record entries, CMS does not require the scribe to sign/date the documentation." The treating physician's signature affirms the note.

That framework was engineered for human scribes. It contains:

• No provenance requirement for the algorithm or model that produced the note.

• No audio integrity verification linking the note to its source recording.

• No cross-jurisdictional consent logic for telehealth encounters crossing state lines.

• No synchronous A/V attestation mechanism to satisfy telehealth parity requirements.

• No guidance on FHIR Provenance resources or HIE interoperability for AI-generated records.

The North Dakota Board of Medicine recognized this gap. Under its 2026 administrative rulemaking—aligned with North Dakota's telehealth parity statutes and informed by the ONC's FHIR interoperability mandates—the Board now requires that any clinical documentation produced by an AI system must include a Source Authenticity Statement to be considered valid for medical practice, audit defense, and insurance reimbursement within the state.

What a Source Authenticity Statement Must Contain

The Source Authenticity Statement is not a disclaimer appended to the bottom of a note. It is a structured provenance artifact with mandatory data elements:

Scribing.io's Cryptographic Provenance Architecture

Scribing.io outputs a cryptographically signed Source Authenticity Statement as part of every encounter. The statement is signed with JSON Web Signature (JWS) and includes every data element above. This artifact is not a PDF appended after the fact—it is generated in real time, hashed against the raw audio, and stored as an immutable record alongside the clinical note.

Critically, because many EHR tenants disable third-party application write access to FHIR R4 Provenance resources, Scribing.io implements a fallback architecture: the Source Authenticity Statement is attached as a DocumentReference resource with JWS envelope and header-level provenance metadata. This ensures the audit trail survives:

• EHR export to external systems

• Health Information Exchange via the North Dakota Health Information Network (NDHIN)

• Payer audit requests and Board of Medicine inquiries

• Patient right-of-access requests under HIPAA's Right of Access rule (45 CFR § 164.524)

No other ambient AI scribe on the market bundles provenance generation, cross-jurisdiction consent logic, and POS/modifier validation into a single, cryptographically verifiable output. This is the compliance hinge that determines whether your AI documentation stands up under scrutiny—or collapses at the first audit.

3. Scribing.io Clinical Logic — Handling a Cross-Border Fargo–Moorhead Telehealth Encounter

This scenario is the centerpiece of Scribing.io's clinical logic and the most common cross-border risk pattern for North Dakota telehealth programs. The Fargo–Moorhead metropolitan area splits across two states, meaning a significant percentage of a Fargo clinician's telehealth panel physically sits in Minnesota during any given visit.

The Scenario

A Fargo-based family medicine physician conducts a scheduled video visit with an established patient who is physically located in Moorhead, Minnesota—just across the Red River. The physician, operating under the assumption that North Dakota's one-party consent rule (N.D. Cent. Code § 12.1-15-02) applies, records the session using an ambient AI scribe without obtaining explicit on-record consent from the patient.

What Goes Wrong Without Scribing.io

1. Privacy complaint filed. Minnesota is also a one-party consent state (Minn. Stat. § 626A.02), but the patient—unfamiliar with the law—files a complaint with both the Minnesota Board of Medical Practice and the North Dakota Board of Medicine, asserting they were never informed that AI was generating their medical record. Under the ND Board's 2026 rule, the absence of a Source Authenticity Statement means the physician cannot prove compliant AI documentation practices.

2. Payer denial of $1,650. The claim is submitted with POS 02 (Telehealth Provided Other than in Patient's Home) instead of POS 10 (Telehealth Provided in Patient's Home), and modifier 95 is missing. The payer denies the claim for missing synchronous attestation and incorrect place-of-service coding.

3. Audit exposure. The AI-generated note contains no Source Authenticity Statement. The North Dakota Board flags the record as non-compliant with 2026 telehealth parity documentation standards. The physician faces a desk audit.

Total exposure: $1,650 in lost revenue + Board inquiry costs + potential corrective action + malpractice insurance notification requirement.

Step-by-Step: How Scribing.io Prevents Every Failure Point

The Granular Logic Flow

1. Pre-encounter geolocation check (T-minus 60 seconds): When the patient joins the video session, Scribing.io resolves the patient's IP address and, where available, caller ID metadata. The system identifies Moorhead, MN—Clay County—as the patient's location. This triggers the cross-border consent engine.

2. Consent law lookup: The engine queries its jurisdiction table. Minnesota (Minn. Stat. § 626A.02) is one-party. However, the ND Board's 2026 Source Authenticity Statement mandate requires documented consent basis regardless of whether the remote jurisdiction is one-party or two-party. The system determines that explicit verbal consent is the safest posture—and the only posture that survives a Board complaint.

3. Real-time consent prompt: The clinician's interface displays a banner: "Patient located in MN. Explicit AI documentation consent recommended. Tap to play consent script." The system plays a configurable verbal disclosure: "This visit is being documented with the assistance of an AI scribe. Your consent is noted in the record. Do you have any questions?" The patient's verbal acknowledgment is timestamped and logged.

4. POS resolution: The patient's resolved address is residential. The system auto-tags POS 10 (Telehealth Provided in Patient's Home) rather than POS 02. This distinction—introduced by CMS's telehealth POS split—is the single most common billing error in cross-border telehealth and the primary reason for the $1,650 denial in this scenario.

5. Modifier 95 validation: Throughout the encounter, the system monitors bidirectional audio-video stream integrity. At session close, it confirms synchronous A/V criteria are met and auto-appends modifier 95 to eligible CPT codes. The jitter and packet-loss metrics are embedded in the Source Authenticity Statement as synchronous proof.

6. Speaker diarization and participant roster: Fourteen minutes into the visit, the patient's spouse asks a question off-camera. Scribing.io's diarization engine detects an unregistered voice. The clinician is prompted: "New speaker detected. Capture consent or redact?" The clinician introduces the spouse; consent is captured; the roster updates to three participants.

7. Source Authenticity Statement generation: At encounter close, the system compiles the full provenance artifact—consent basis (cross-border explicit, MN), audio hash (SHA-256), model version (Scribing.io Clinical v4.2.1), NPI, prompt class (SOAP—Family Medicine—Follow-up), device fingerprint, POS 10, modifier 95, and three-member participant roster. The statement is JWS-signed and written to the EHR via FHIR DocumentReference with provenance headers.

8. Claim pays first pass. The Source Authenticity Statement, correct POS, and modifier 95 ensure clean claim submission. No denial. No appeal. No revenue leakage.

9. Complaint closed. When the Board inquiry arrives, the practice exports the audit packet—Source Authenticity Statement, timestamped consent recording, participant roster, and audio hash verification—directly from Scribing.io. The complaint is closed administratively.

4. CMS Signature Authentication — Why Transmittal 713 Does Not Cover AI Scribes

CMS Transmittal 713 (CR 10076), codified in Pub 100-08, Chapter 3, §3.3.2.4, governs how scribe-produced documentation is authenticated. The operative sentence: the treating physician or NPP must "sign and date" the medical record entry, and their signature serves as attestation that the information is accurate and complete.

The gap: Transmittal 713 assumes the scribe is a human being in the room or on the call. It was never updated for AI-generated documentation, which introduces risks that human scribe workflows do not:

• Hallucination risk: An AI model may generate clinically plausible but factually incorrect content. A physician's signature on an AI-generated note must therefore be informed by awareness that the note was AI-produced—not just a passive co-sign. The JAMA perspective on AI documentation integrity emphasizes that physician attestation carries heightened responsibility when the source is algorithmic.

• Version drift: Unlike a human scribe whose competence is relatively stable, an AI model can change behavior between versions. Without model version tracking in the record, an audit cannot determine whether the note was produced by a validated or deprecated model.

• Provenance gap: A human scribe's identity is verifiable through HR records and credentialing. An AI scribe's identity requires explicit metadata—model name, version, prompt class, and audio hash—none of which Transmittal 713 mandates.

Scribing.io closes this gap by embedding model provenance directly into the clinical note and the Source Authenticity Statement. The physician's signature, when applied via standard EHR attestation workflow, sits atop a complete provenance chain—not a black box.

5. Technical Reference: ICD-10 Documentation Standards

AI-generated clinical documentation frequently under-specifies diagnosis codes, defaulting to unspecified or generic codes that trigger payer medical review and reduce reimbursement. This is not a theoretical risk—a 2023 NIH-indexed study on AI clinical documentation found that automated systems routinely selected lower-specificity ICD-10 codes when encounter context was ambiguous.

Scribing.io's clinical engine addresses this through specificity escalation logic: the system analyzes the full encounter transcript—chief complaint, history, review of systems, assessment, and plan—and maps findings to the highest-specificity ICD-10-CM code supported by documented evidence. The system will not assign a code that lacks transcript support, but it will flag when documentation supports a more specific code than the one a physician might select manually.

Common Telehealth Encounter Codes

Two codes appear with high frequency in North Dakota telehealth encounters, particularly in administrative and counseling contexts:

• Z02.89 — Encounter for other administrative examinations; Z71.89 — Other specified counseling

Z02.89 is commonly used for pre-employment physicals, fitness-for-duty evaluations, and insurance examinations conducted via telehealth. The denial risk: payers flag Z02.89 claims that lack documentation of the specific administrative purpose. Scribing.io's engine extracts the stated purpose from the encounter transcript and populates the note's "Reason for Visit" field with the specific administrative context, preventing the vague documentation that triggers medical review.

Z71.89 covers counseling encounters that do not fit into more specific Z71 subcategories—dietary counseling (Z71.3), alcohol abuse counseling (Z71.41), or substance abuse counseling (Z71.51), for example. The risk here is under-coding: a physician bills Z71.89 when the transcript clearly documents dietary counseling that should be coded as Z71.3. Scribing.io's specificity escalation logic detects this pattern and suggests the more specific code, with transcript evidence highlighted for the physician's review.

Specificity Escalation in Practice

Every code suggestion includes a direct link to the supporting transcript segment, allowing the physician to verify specificity before signing. This workflow satisfies both the AMA's CPT documentation principles and the ND Board's requirement that AI-generated content be physician-verified before becoming part of the legal medical record.

6. Telehealth Parity Compliance — POS Codes, Modifier 95, and Real-Time Validation

North Dakota's 2026 telehealth parity statutes require that synchronous audio-video telehealth encounters be reimbursed at parity with in-person visits—provided the documentation meets specific technical requirements. Two coding elements are the primary failure points:

Place of Service: POS 02 vs. POS 10

• POS 02 — Telehealth Provided Other than in Patient's Home. Used when the patient is at a clinic, hospital, or other healthcare facility during the telehealth encounter.

• POS 10 — Telehealth Provided in Patient's Home. Used when the patient connects from their residence.

The distinction matters for reimbursement. CMS applies facility pricing to POS 02 claims and non-facility pricing to POS 10 claims. Submitting POS 02 when the patient is at home results in lower reimbursement. Submitting POS 10 when the patient is at a facility triggers audit flags. Scribing.io resolves patient location at session start and auto-tags the correct POS code—eliminating the manual lookup that billers routinely get wrong in cross-border encounters.

Modifier 95: Synchronous Telehealth

Modifier 95 indicates that the service was rendered via real-time interactive audio-video telecommunications. Its absence on a telehealth claim is the most common reason for parity denial. Scribing.io validates synchronous A/V continuity throughout the encounter by monitoring stream metrics. If the video feed drops below the CMS-required interactive threshold (e.g., audio-only for more than 50% of the encounter), the system flags the encounter as potentially audio-only and suggests the appropriate audio-only modifier instead, preventing a modifier 95 claim that would fail on audit.

7. Implementation Checklist for North Dakota Medical Directors

Deploying AI scribing in a North Dakota telehealth program requires coordinating legal, clinical, technical, and billing workflows. This checklist distills the playbook into executable steps.

1. Verify BAA coverage. Confirm that your AI scribe vendor has executed a Business Associate Agreement that specifically covers ambient audio capture, AI processing, and storage of PHI. Scribing.io's BAA is pre-configured for HIPAA 2026 requirements including the Security Rule's updated AI provisions.

2. Configure cross-border consent rules. For Fargo–Moorhead and other border-area practices, enable Scribing.io's geolocation-based consent engine. Set your default posture: we recommend explicit verbal consent for all cross-state encounters regardless of the remote state's consent classification.

3. Enable Source Authenticity Statement generation. This is on by default in Scribing.io. Verify that your EHR integration is writing the statement as a DocumentReference with JWS headers. If your EHR supports FHIR R4 Provenance write access, enable the primary pathway for richer interoperability.

4. Train physicians on attestation responsibility. Reinforce that signing an AI-generated note carries the same legal weight as signing a human-scribed note. Provide the AMA's AI attestation guidance as a reference. Scribing.io highlights AI-generated content in the note with visual markers to support informed review.

5. Audit POS and modifier accuracy. Run a retrospective analysis of your last 90 days of telehealth claims. Identify encounters coded POS 02 where the patient was at home (should be POS 10) and encounters missing modifier 95. Scribing.io's analytics dashboard surfaces these patterns automatically.

6. Establish an export protocol for Board inquiries. Pre-configure your audit packet export: Source Authenticity Statement, timestamped consent recording, participant roster, SHA-256 audio hash, and physician attestation timestamp. Scribing.io bundles these into a single downloadable archive.

7. Validate NDHIN interoperability. If your practice participates in the North Dakota Health Information Network, confirm that AI-generated records with Source Authenticity Statements pass NDHIN's validation rules. Scribing.io's DocumentReference format is pre-validated against NDHIN's 2026 schema requirements.

8. Review ICD-10 specificity patterns. Enable Scribing.io's specificity escalation alerts. Audit a sample of 50 recent AI-generated notes to verify that codes reflect maximum supported specificity. Pay particular attention to Z-code encounters (Z02.89, Z71.89) where under-specification is most common.

Ready to Operationalize This Playbook?

Book a 15-minute demo to see our ND Source Authenticity Statement generator + cross-border consent engine with EHR FHIR Provenance/DocumentReference fallback and real-time modifier 95/POS 02–10 validator—ready for 2026 parity audits.

The question is no longer whether AI scribing is legal in North Dakota. It is. The question is whether your AI scribe produces documentation that survives the Board's Source Authenticity Statement mandate, CMS signature requirements, telehealth parity coding rules, and cross-border consent exposure—all at once, on every encounter, without manual intervention. That is the operational standard Scribing.io was built to meet.