Every AI rule that will govern your practice for the next decade is being drafted right now, in a public meeting room with an empty chair at the table. The California BBS, New York's Office of the Professions, the Texas Behavioral Health Executive Council, Florida's DOH Chapter 491 boards, and Illinois IDFPR are writing, or about to write, the rules on AI-assisted documentation, supervision of AI use by associates, and clinician accountability for machine-drafted records. So are PSYPACT, the Counseling Compact, and the Social Work Compact. So are HHS, SAMHSA, and CMS. Clinicians who show up to comment will shape rules they can live with; clinicians who wait will inherit rules written by vendors, payers, and people who have never sat with a client at risk. This lesson teaches the machinery of rulemaking participation and how to write a comment letter a board actually uses. By the end you will have drafted a state-board comment letter on a live AI rule, ready to file.

The Empty Chair at the Drafting Table

Carry this image through the lesson: every regulation is written at a table, and there is a chair at that table with your profession's name on it. The chair does not stay empty. If a licensed clinician who runs AI-assisted documentation under a real caseload does not sit in it, someone else will: a scribe vendor's government-affairs lead, a payer's policy analyst, a board staffer who has never reviewed a 90837 note, or a legislator's aide whose mental model of AI in therapy comes from a news story about a chatbot. None of them knows what happens at 9:54 PM when Maria, the Oakland LCSW, has seven half-finished notes and a scribe draft she has not yet verified. None of them has signed an attestation. The rule that emerges from a table without a clinician is not malicious; it is written by people who do not know what they do not know, and you will practice under it for ten years.

You have already seen what happens when the chair sits empty. The Illinois WOPR Act (2025) drew a national bright line against AI providing therapy, and its 2026 enforcement posture is still being worked out in guidance clinicians can influence. Nevada AB 406 (2025) regulated AI in mental and behavioral healthcare; New York passed AI companion safeguards in 2025; Colorado is running a dual track between the original AI Act (SB 24-205, postponed effective date June 30, 2026) and SB 26-189, the May 2026 ADMT-focused replacement effective January 1, 2027 if signed, with mental-health bills HB 26-1139 and HB 26-1195 still in play. Every one of those laws went through a comment-and-hearing process, and the texts show where clinicians testified: the statutes that distinguish AI drafting documentation from AI purporting to provide care are the ones practicing clinicians helped write. This is the Level 5 step after the policy, the pilots, and the audits: you stop adapting to rules and start shaping them. Comment on proposed rules; do not wait for them.

The Five State Boards and How Each One Makes Rules

Start with the five boards this program tracks, because each has a different door you walk through. The California Board of Behavioral Sciences (BBS) licenses LMFTs, LPCCs, and LCSWs and makes rules through public meetings with posted agendas and written-comment windows; it is the most consequential behavioral health board in the country by licensee count, and its policy committee is where AI documentation and AI-in-supervision questions surface first. New York's Office of the Professions (OPP) regulates mental health practitioners through the state register's formal comment periods, and New York's 2025 AI companion safeguards law guarantees that OPP guidance on clinician AI use is coming. The Texas Behavioral Health Executive Council (BHEC) is an umbrella council over the psychology, counseling, social work, and MFT boards, so a single well-placed comment can shape rules across four professions. Florida regulates LMFTs, LMHCs, and LCSWs under Department of Health Chapter 491, with rule workshops open to public participation. And Illinois IDFPR is the most active jurisdiction on clinician AI because the WOPR Act handed it an enforcement mandate; its guidance cycle is where the line between "AI-assisted documentation" and "AI providing therapeutic services" gets drawn in operational detail.

The mechanics matter because participation is procedural, not rhetorical. Every board publishes agendas in advance; every formal rulemaking includes a written-comment window, usually 30 to 45 days; comments received inside the window become part of the administrative record the board must generally address, while a comment filed after the window closes is a letter to the editor. And honor the state-specific discipline from every compliance lesson in this program: never generalize across jurisdictions. A comment to the BBS cites California authority (the BBS's own regulations, CA Civ Code §43.92 on duty to protect, CA SB 855 on parity where relevant); a comment to IDFPR engages the WOPR Act's actual text. A letter that cut-and-pastes generic talking points across five boards tells each board you do not know its law, and boards discount commenters who do not know their law.

The Compacts: Where One Rule Crosses Forty Borders

The interstate compacts are the highest-leverage and least-attended tables in behavioral health governance. PSYPACT governs interjurisdictional psychology practice; the Counseling Compact and the Social Work Compact do the same for counselors and social workers. Each is administered by a commission of state delegates that adopts rules binding across every member state, which means a single compact rule on AI-assisted documentation in cross-state telehealth applies simultaneously in dozens of jurisdictions. Commenting on one compact rule can be worth commenting on forty state rules.

The AI questions the compacts must answer are concrete, unresolved, and exactly the questions a practicing clinician is positioned to answer well. When a PSYPACT psychologist uses an AI scribe to document a telehealth session with a client in another state, whose AI rules apply: the home state's, the client's state's, or a compact standard? If the client sits in Illinois, does the WOPR Act reach a compact-practicing clinician licensed elsewhere? Should compact rules require AI-use disclosure in the cross-state informed consent? Right now those questions are being answered by default, by silence, or by whoever attends the commission meetings. Compact commissions publish proposed rules and take written comment exactly as state boards do; the difference is that the audience is a fraction of the size, so a substantive clinician comment carries far more weight per page.

The strategic posture: treat compact rulemaking as the early-warning system. Compact rules crystallize before most member states address the same question, and member-state boards then borrow compact language when drafting their own. The empty chair at the compact table is the emptiest chair in the field, and it governs the most territory.

Comment on proposed rules; do not wait for them. The rule will exist either way. The only question your participation decides is whether it was written by someone who has ever signed a note.

The Federal Layer: HHS, SAMHSA, and CMS Comment Cycles

Above the boards and compacts sits the federal layer, running the same machinery at larger scale. HHS, through the Office for Civil Rights, owns HIPAA, including the psychotherapy-notes carve-out at 45 CFR 164.508(a)(2) and the §164.501 definition; how ambient session capture interacts with that protection is precisely where clinician comments change final-rule text. SAMHSA owns 42 CFR Part 2; the 2024 final rule that aligned Part 2 with HIPAA on single consent and redisclosure was itself the product of a comment cycle, and the next round of Part 2 guidance will decide how AI vendors and their subprocessors fit inside Part 2 programs, a question every SUD clinician using a scribe needs answered in writing. CMS sets the Medicare rules that ripple through the entire payer market: the 2024 expansion that brought LMFTs and LMHCs into Medicare Part B went through rulemaking, the annual Physician Fee Schedule proposed rule is a standing yearly comment opportunity on documentation standards and the 90791 and 90832/90834/90837 code family, and CMS positions on AI-drafted records will shape what every commercial payer demands.

The comment mechanics are clinician-friendly once demystified. Proposed federal rules publish in the Federal Register with a comment deadline, typically 60 days; agencies must respond to significant comments in the final rule's preamble, and a well-argued comment the agency cannot rebut frequently changes final text. The MHPAEA saga is the standing proof that federal posture is unstable and therefore influenceable: the 2024 NQTL final rule was followed by the May 2025 federal non-enforcement statement and the March 2026 court-filing disclosure that the Departments will propose replacement regulations rather than defend the September 2024 rule. Replacement regulations mean a new comment cycle; clinicians who treat it as someone else's job will get a parity framework written without them, while statutory MHPAEA rights and the CAA 2021 comparative-analysis requirements persist and state parity laws (NY Timothy's Law, the Illinois parity statute, CA SB 855 with DMHC regulations) remain enforceable regardless. One discipline keeps federal commenting tractable: comment only where your direct practice experience is evidence the agency cannot get elsewhere. Opinions are abundant in a docket; operational evidence from a licensed clinician under a real caseload is scarce, and scarcity is what agencies quote.

What Boards Actually Need to Hear: The Three Clinician-Only Contributions

Regulators drafting AI rules consistently lack three inputs only practicing clinicians can supply, and an effective comment letter delivers at least one. First, the workflow ground truth. Boards drafting in the abstract conflate categories your practice separates cleanly: an AI scribe that drafts a SOAP note for clinician verification and signature is not a chatbot purporting to deliver therapy, and a rule written without that distinction either bans legitimate documentation support or blesses dangerous care substitutes. Your comment supplies the taxonomy: drafting versus clinical decision-making; the signature as the legal attestation on every record; the verification pass before signing as the control point a rule should regulate. The line this entire program holds is the line you hand the board: AI never scores the CSSRS, never assigns a risk level, never makes the Tarasoff or duty-to-protect determination, never makes the mandated-report call; the clinician decides, and AI structures and formats after the clinical determination. A board that writes that allocation of responsibility into rule has written a rule clinicians can comply with.

Second, the foreseeable-consequence map. Regulators rarely see a rule's second-order effects on access and on pre-licensed clinicians. A rule requiring written client consent before any AI touches a record is defensible; a rule drafted so broadly it requires re-consent for spell-check inside the EHR is not, and you are the person who can show the board the difference. A supervision rule that silently invalidates an associate's hours, the Carmen problem from this program's opening, where an AMFT's disclosed scribe use suddenly threatens whether her supervised hours count, is the kind of consequence a board will fix in drafting if a commenter shows it the scenario, and will spend three years fixing in waiver requests if no one does. Third, the enforceability test. Boards must enforce what they write: is the rule auditable from the record (a consent addendum, a signed attestation, a vendor BAA on file), or does it require the board to adjudicate what happened inside a model? Rules that attach to artifacts survive; rules that attach to intentions do not.

The Anatomy of a Comment Letter a Board Actually Uses

Most comment letters fail the same way: they argue a position instead of improving a text. Board staff sort comments into two piles, "supports/opposes" and "proposes specific changes," and only the second pile changes language. The letter you will build has five parts designed to land in the second pile. Part one, standing: who you are, in two sentences establishing why your evidence is scarce: license type, years in practice, setting, caseload, AI-relevant experience ("I have operated an AI-assisted documentation program across a 25-clinician group practice for two years under a written policy, BAA review, and supervision integration"). Part two, the specific provision: quote the exact section and language you are addressing. A letter about "the proposed rule" generally is a letter about nothing; a letter about "proposed §X(b)(2)'s phrase 'any use of artificial intelligence in the provision of services'" is actionable.

Part three, the operational consequence: one concrete, worked scenario showing what the language does in a real practice. Show the 53-minute psychotherapy session, the scribe draft, the verification pass, the signature; show where the proposed language helps and exactly where it misfires, using real artifacts: the consent addendum you actually use, the supervision-contract clause, the pre-signature verification step. Part four, proposed alternative text: the single highest-leverage element of any comment. Draft the replacement sentence yourself, in regulatory register, attached to an auditable artifact; boards adopt proposed text far more often than commenters expect, because staff drafting under deadline will take a well-crafted sentence that solves the problem they now see. Part five, the offer: make yourself available for the workshop, the testimony, the stakeholder panel; the second invitation is where standing influence begins.

Two style rules govern the whole letter. Keep it under two pages; a board staffer summarizes forty comments in an afternoon, and the letter that summarizes itself gets quoted. And concede the legitimate concern before you propose the fix: a letter that pretends AI poses no documentation risk reads as vendor copy, while a letter that names the hallucination risk, the consent gap, and the supervision exposure, then proposes the text that addresses them, reads as the profession governing itself.

Building the Rulemaking Calendar and the Coalition Behind the Letter

Influence is logistics before it is eloquence. The first piece of infrastructure is a rulemaking calendar: a quarterly-reviewed tracker of every body whose rules reach your practice, recording where proposed rules publish, the comment-window length, the next meeting, and the AI items pending. For a California group practice with compact-credentialed clinicians and Medicare patients, the tracker covers at minimum the BBS policy committee; IDFPR if any clinician serves Illinois clients; the relevant compact commissions; and the federal trio, with the CMS Physician Fee Schedule as the guaranteed annual entry and the anticipated MHPAEA replacement regulations as the watched item. Thirty minutes a quarter maintains the calendar. Missing one 30-day window can cost a decade of practicing under a sentence you could have fixed.

The second piece is the coalition. Your professional association chapters (the state affiliates of APA, NASW, AAMFT, ACA, and for behavior analysts the structures around the BACB) maintain government-affairs functions that monitor these dockets; channeling your letter through a chapter multiplies its standing. Coordinated letters making different arguments outperform identical form letters, which boards count as one comment: assign the workflow letter to the practicing clinician, the supervision-consequences letter to the board-approved supervisor, the enforcement-artifact letter to the compliance officer. The third piece is the follow-through that converts a comment into a relationship: attend the meeting where the item is heard, give the two-minute public comment referencing your written submission, and apply when the board forms an AI stakeholder workgroup. The progression is predictable, comment letter, public testimony, workgroup member, committee appointee, and each step is open to any licensee willing to do the unglamorous reading. The field's AI rules for the 2030s will be written by the clinicians who started filing comments in the 2020s.

The Applied Problem: Your State-Board Comment Letter on a Live AI Rule

Your artifact is a filed-ready State-Board AI Comment Letter, drafted against a live or modeled proposal. If your board has an open AI item, use it; if not, use this modeled provision, which mirrors language circulating in several jurisdictions: "Proposed Rule §1820.5(b): A licensee shall not use artificial intelligence in the provision of behavioral health services without the prior written approval of the Board." Your task is the two-page letter that turns this well-intended, unworkable sentence into a rule that protects clients without criminalizing the verified scribe draft.

Step one: draft parts one and two. Write your standing paragraph (license, setting, caseload, AI-program experience) and quote the provision exactly, naming the operative defect: "in the provision of behavioral health services" fails to distinguish documentation support under clinician attestation from AI purporting to deliver care, the distinction the Illinois WOPR Act drew. Step two: write the operational-consequence scenario from your own week: the 90837 session, the ambient draft, the verification pass under the cardinal rule (the clinician signs the note; the signature is a legal attestation), the risk boundary stated explicitly (AI never scores the CSSRS, never makes the duty-to-protect determination; the clinician decides and AI formats after). Then show the misfire: the proposed text makes this defensible workflow require individualized board pre-approval the board cannot staff, while a direct-to-consumer chatbot outside licensure escapes the rule entirely.

Step three: draft the alternative text, attached to auditable artifacts. A strong version: "A licensee may use artificial intelligence tools to assist with documentation and administrative tasks provided that (1) the client has consented through a signed AI-use addendum retained in the record; (2) the licensee reviews and verifies every AI-assisted record before signing, and the licensee's signature constitutes attestation to its accuracy; (3) no artificial intelligence tool performs clinical assessment, risk determination, diagnosis, or treatment decisions, which remain the sole responsibility of the licensee; and (4) for pre-licensed registrants, AI use is disclosed in the supervision agreement and overseen by the supervisor of record." Every clause attaches to an artifact an investigator can request. Step four: close with the offer to testify, then run the verification pass: under two pages; quotes the exact provision; concedes the legitimate risk before fixing it; proposes adoptable text. Done looks like: a dated, signed, two-page letter with all five parts, alternative text a board staffer could paste into a draft, and a calendar entry for the comment deadline and the meeting where the item will be heard.

Key Takeaways

• Every AI rule governing behavioral health practice is drafted at a table with an empty chair reserved for a practicing clinician. If you do not fill it, vendors, payers, and staff who have never signed a note will. The operating instruction: comment on proposed rules; do not wait for them.
• The five boards, CA BBS, NY OPP, TX BHEC, FL DOH Chapter 491, and IL IDFPR, make rules through posted agendas and written-comment windows; comments filed inside the window enter the administrative record the board must address. BHEC's umbrella structure means one comment can shape four professions; IDFPR's WOPR enforcement guidance is the live national front line.
• Compact rulemaking (PSYPACT, Counseling Compact, Social Work Compact) is the highest-leverage, least-attended table: one commission rule binds across every member state, compact language pre-shapes the state rules that follow, and a substantive clinician comment carries outsized weight because so few are filed.
• The federal layer runs on notice-and-comment at scale: HHS/OCR on HIPAA and the psychotherapy-notes carve-out at 45 CFR 164.508(a)(2), SAMHSA on 42 CFR Part 2 under the 2024 final rule, CMS on Medicare behavioral health and the annual fee-schedule cycle. The MHPAEA trajectory guarantees a parity comment cycle clinicians must not skip, while state parity laws like CA SB 855 and NY Timothy's Law remain enforceable throughout.
• Boards need three things only clinicians can supply: the workflow ground truth separating documentation support from AI purporting to provide care (AI never scores the CSSRS, never makes the risk, Tarasoff, or mandated-report determination); the foreseeable-consequence map, including effects on associates' supervised hours; and the enforceability test that rules must attach to auditable artifacts, not intentions.
• An effective comment letter has five parts: standing, the exact quoted provision, one worked operational scenario, proposed alternative regulatory text attached to artifacts an investigator can request, and an offer to testify. Under two pages, conceding the legitimate risk before proposing the fix.
• Influence is logistics: a quarterly rulemaking calendar across boards, compacts, and the federal trio; coordinated (not identical) letters channeled through association chapters; and the ladder from comment to testimony to workgroup to committee seat. The clinicians filing comments in the 2020s will hold the chairs in the 2030s.