Somewhere in the next few years, a sentence about artificial intelligence will be written into the APA Code of Ethics, the NASW Code, the AAMFT Code, the ACA Code, and the BACB Ethics Code for Behavior Analysts, and from that day forward every clinician in your profession will be disciplinable against it. That sentence is being drafted now, in committee rooms most clinicians never think about, by the small number of practitioners who volunteered for the unglamorous work of standards revision. This lesson is about becoming one of them. You have spent four levels learning to practice under codes; the Level 5 move is to help write the next one, so that the AI provisions your profession lives under for the next decade reflect how clinical AI actually works: the clinician deciding, the machine formatting, the signature attesting. By the end, you will have built a Standards-Committee Submission Outline, the position-to-provision document that turns your practice experience into proposed code language a revision committee can adopt.

The Constitutional Convention of Your Profession

Carry this analogy through the lesson: an ethics-code revision is the constitutional convention of a profession. Journal articles, conference panels, white papers, and position statements are the pamphlets and the public debate, important, influential, and not binding on anyone. The code is the constitution: the document a licensing board cites in a disciplinary order, a malpractice attorney reads into a deposition, a supervisor incorporates into a supervision contract, and an ethics committee applies to a complaint. The distance between pamphlet and constitution is the committee room. A position paper, however brilliant, changes nothing until someone carries its ideas into a code-revision process and converts them into provision language that survives drafting, comment, and adoption. The playbook states the mission in one line: serve on the standards-body committees that turn position papers into binding codes.

This distinction matters enormously for AI because the field is currently drowning in pamphlets. Every association has issued guidance, every journal has run special issues, every conference has an AI track, and almost none of it binds. A clinician who violates a white paper has violated nothing; a clinician who violates the code faces the ethics committee and, because state boards incorporate professional codes into their regulations, often the board as well. When the next revision cycle opens, the AI question moves from commentary to canon, and the people in the room decide whether the canon is written by clinicians who have run AI-assisted documentation under real caseloads or by members whose closest contact with a scribe is a vendor demo. You have seen this dynamic at the state boards and at the payer table; the standards body is the third table and the most durable: regulations get rewritten every few years, contracts every cycle, but a code revision can stand for a decade and shapes all the others, because boards, courts, and payers all read the code.

One more reason this work belongs to you specifically: codes are written by volunteers. There is no standards-drafting profession; there are practitioners who answered a call for committee members, reviewers, and commenters. The barrier to the room is not credentials beyond your license and standing; it is willingness to do slow, careful, uncompensated drafting work over multiple years. That is precisely why the room is underpopulated by working clinicians and why your seat is available.

Five Codes, Five Cultures: Knowing the Document You Want to Change

You cannot influence a code you have not studied as a drafter studies it, and the five codes this program tracks are different documents with different revision cultures. The APA's Ethical Principles of Psychologists and Code of Conduct distinguishes aspirational principles from enforceable standards, and the enforceable standards are written with near-statutory precision because they are litigated; an AI proposal to APA must be drafted at that precision. The NASW Code of Ethics carries a strong values architecture, service, social justice, dignity and worth of the person, and its informed-consent and privacy provisions (the territory of NASW 1.07) are where AI questions naturally land; a proposal framed in the code's own values language travels farther than one framed in technology language. The AAMFT Code of Ethics governs a relational profession; its consent provisions (the AAMFT 1.6 territory) must handle multiple clients in the room, which makes AI consent provisions for couples and family work, who consented to the recording, whose data is in the transcript, materially harder and squarely in need of practitioner drafting. The ACA Code of Ethics already contains a distinct section on distance counseling, technology, and social media (Section H), which means the ACA revision question is often amendment of existing technology provisions rather than creation from nothing; knowing the existing section's structure is the price of admission. And the BACB Ethics Code for Behavior Analysts (2022) operates in the most enforcement-forward culture of the five, with certification consequences and a Medicaid-heavy, documentation-intensive practice base, ABA documentation under CPT 97151 through 97158, RBT supervision logs, where AI documentation provisions must fit an existing machinery of behavior-analytic precision.

The shared inheritance across all five is the consent-primacy principle this program met in the EAP-platform lesson: APA Standard 4's confidentiality architecture, NASW 1.07, AAMFT 1.6, and ACA Section H all require the clinician to obtain meaningful informed consent in the clinician's own voice, and the standards question AI now poses is how that obligation operates when platforms pre-consent members, when scribes capture ambient audio, and when subprocessors sit behind the vendor. Boards have already begun answering pieces of this, the playbook notes that CA BBS, NY OPP, and TX BHEC have addressed contractor-platform ethics primacy in recent guidance, and the codes will consolidate the answer. The drafter who arrives knowing each code's existing consent provisions, by number, in the code's own vocabulary, is the drafter whose proposed language gets read.

How a Code Actually Changes: The Revision Machinery

Code revision runs on machinery you can map and enter, and it rhymes with the rulemaking machinery from the state-board lesson. A revision cycle typically begins when the association's governance charges a revision task force or ethics-code committee. The committee gathers input: member surveys, public comment periods, written submissions from members and committees, subject-matter consultations. Drafts circulate, sometimes publicly, for comment. The final text goes to the association's governing body for adoption. Three properties of this machinery determine your strategy. First, it is episodic: revision cycles open years apart, which means the preparation happens before the cycle opens, and the clinician with a submission ready on day one of the comment window outruns the clinician who starts drafting when the draft is published. Second, it is relational: committees are staffed from the association's existing committee ecosystem, ethics committees, technology committees, state and division leadership, so the path to the revision table runs through years of smaller service, the state chapter ethics committee, the division task force, the reviewer pool. Third, it is textual: like the board comment letter, the submission that proposes specific provision language in the code's own register beats the submission that argues a position, because committee drafters under deadline adopt usable sentences.

There is also a between-cycles channel that most members never use: ethics committees issue advisory opinions and interpretive guidance continuously, and associations publish practice standards and guidelines between code revisions. A well-built submission to an ethics committee asking for interpretive guidance on, say, whether ambient AI capture requires consent renewal mid-episode, does two things: it gets working clinicians an answer years before the next code revision, and it seeds the precedent the revision committee will later consolidate. Interpretive guidance is the case law of a code; revisions tend to codify what the opinions have already worked out. The clinician who shapes the opinions has pre-drafted the revision.

A position paper persuades the people who already agree with it. A code provision binds the people who never read the paper. The committee room is where one becomes the other.

What the AI Provisions Must Say: The Clinician-Drafter's Substantive Agenda

Walk into the committee with a substantive agenda, not just availability, and let the agenda be the one this entire program has built. Provision one: responsibility allocation. The code must state that the use of AI tools does not diminish the clinician's professional responsibility for any clinical record, assessment, or communication; the clinician who signs attests, and "the AI drafted it" is not a defense before an ethics committee. This codifies the cardinal rule, the signature is a legal and ethical attestation, not a formatting step, into binding language. Provision two: the clinical-determination boundary. The code must reserve clinical judgment to the clinician in enforceable terms: no AI system performs assessment, diagnosis, risk determination, or treatment decision-making within the professional relationship. In this program's hard formulation: 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; AI structures, transcribes, and formats after the clinician's determination. Code language that draws this boundary gives every future ethics committee a clean adjudication line and gives every honest vendor a design constraint.

Provision three: informed consent in the clinician's voice. The code must require disclosure and consent for AI participation in services and records, obtained by the clinician, in the clinician's own consent process, with platform or employer pre-consent explicitly insufficient to discharge the clinician's obligation. This is the EAP-platform lesson written into canon: APA Standard 4, NASW 1.07, AAMFT 1.6, and ACA Section H already carry the consent-primacy DNA, and the AI provision extends it to the ambient-capture era, including the client's right to decline AI participation without losing access to care. Provision four: competence. Every code carries a boundaries-of-competence obligation; the AI provision extends it, a clinician who uses AI tools must understand their limitations, failure modes, and data handling well enough to supervise their output, and supervisors who oversee AI-using supervisees carry the same obligation one level up. Provision five: supervision and the pre-licensed clinician. The Carmen problem belongs in the code: AI use by supervisees must be disclosed within supervision, and supervisors must address AI use in the supervision agreement. Each provision attaches to an auditable artifact, the signed record, the consent document, the supervision agreement, which is the same enforceability discipline the board-comment lesson taught, because ethics committees, like boards, can only adjudicate what leaves a trace.

Earning the Seat: The Service Ladder From Member to Drafter

Standards influence is earned through a service ladder climbable on a working clinician's schedule. Rung one: membership with presence, join the association's ethics and technology interest communities, attend the meetings where volunteer calls are issued, and respond in writing to every comment opportunity; written comments are how the ethics staff learns your name. Rung two: state and division service, the state chapter's ethics committee or the relevant division's technology task force, where the workload is bounded and the people who staff national committees watch for reliable contributors. Rung three: the reviewer and consultant pool, when the association seeks subject-matter input on AI guidance or practice standards, the clinicians who have already commented well get asked. Rung four: the national committee or task force appointment, by application or nomination when a charge is issued. Rung five: the revision committee itself. The climb takes years, which is exactly why it must begin before the cycle opens, and why the lesson's artifact is built now rather than later.

Two disciplines make the climb productive rather than ornamental. First, bring the data only a practitioner has. Committee deliberations on AI chronically lack ground truth: what an ambient scribe actually mis-hears in a couples session, what a supervisee actually discloses about tool use, what a consent conversation actually sounds like with a trauma client who asks where the recording goes. Your practice, run under the policies this program taught you to build, generates exactly this evidence, de-identified vignettes, error logs from your verification passes, consent-conversation scripts that worked and failed. A committee member who supplies ground truth becomes load-bearing; a committee member who supplies opinions is one vote. Second, respect the difference between your interest and the profession's. If you consult for a vendor, hold equity, or speak for a platform, disclose it and recuse where the conflict-of-interest norms require; standards work runs on trust, and a drafter discovered shading language toward a commercial interest loses the room permanently. The BCBA serving on BACB-adjacent work, the LCSW in NASW service, the LMFT at AAMFT, the counselor at ACA, the psychologist at APA: each carries the same fiduciary posture, the code belongs to the profession and the public it protects, not to the drafter's practice model.

Writing Language Committees Adopt: From Position to Provision

The craft skill of standards work is converting a position into a provision, and the conversion has rules. A position says what you believe: "clinicians should remain responsible for AI-drafted records." A provision says what a member must do, in language an ethics committee can adjudicate: "Clinicians who use artificial intelligence or automated tools in creating clinical records remain fully responsible for the accuracy and content of those records, and shall review each record before signing or otherwise attesting to it." Notice the conversion moves: the actor is identified (clinicians who use), the obligation is behavioral and observable (review before signing), the responsibility is total and non-delegable (remain fully responsible), and the trigger artifact exists (the signed record). Run every proposed sentence through the adjudication test: could a five-member ethics committee, looking at a complaint file, determine whether this sentence was violated? "Use AI responsibly" fails the test. "Obtain the client's informed consent, documented in the record, before employing tools that capture or process session content" passes it.

Three further drafting rules keep your language adoptable. Write technology-neutral where possible: codes outlive products, so "automated tools that record, transcribe, or generate clinical content" survives the decade in a way that any vendor category will not. Draft the duty, not the tool: the obligation should attach to the clinician's conduct (disclose, verify, supervise, remain responsible) rather than to a machine's properties, because conduct is what ethics committees regulate. And preserve the aspirational-versus-enforceable distinction of the target code: APA's enforceable standards demand statutory tightness; a values-forward preamble sentence about technology and human dignity belongs in the aspirational layer, and confusing the layers marks the submission as amateur. Finally, anticipate the committee's objections in the submission itself: the access objection (will this burden under-resourced practitioners?), the innovation objection (does this freeze a moving field?), and the enforceability objection (can a committee actually adjudicate it?). A submission that raises and answers its own objections reads like committee work product, which is exactly what gets absorbed into committee work product.

The Applied Problem: Your Standards-Committee Submission Outline

Your artifact is the Standards-Committee Submission Outline: the position-to-provision document you will adapt and file when your association's revision cycle, comment window, or ethics-committee guidance process opens, and which doubles as your application portfolio for committee service. Build it in four steps. Step one: choose your code and map its anchors. Pick the code that governs your license (APA, NASW, AAMFT, ACA, or BACB) and list the existing provisions your AI proposals attach to: the consent provisions (the Standard 4 / 1.07 / 1.6 / Section H territory), the competence provisions, the records and documentation provisions, the supervision provisions. Your submission must cite these by number and quote their current language, because revision is amendment, and amendment begins from the existing text.

Step two: draft the five provisions in adoptable register. Write your version of the five-part agenda, responsibility allocation, the clinical-determination boundary, consent in the clinician's voice with platform pre-consent insufficient, competence extended to AI oversight, and supervision disclosure, each as one or two sentences that pass the adjudication test, each attached to an auditable artifact (the signed record, the documented consent, the supervision agreement). For the clinical-determination boundary, draft it in your code's vocabulary but keep this program's hard floor intact: no AI assessment, diagnosis, risk determination, or treatment decision; the clinician decides, AI formats after. Step three: attach the ground truth. For each provision, add one de-identified practice vignette showing why the provision is needed and workable: the verification pass that caught a scribe error before signature; the consent conversation in which a client asked where the audio goes; the supervision meeting where a supervisee's undisclosed tool use surfaced. Then add the objection paragraph: answer access, innovation, and enforceability for your package in three or four sentences. Step four: run the verification pass and route it. Check every cited provision number against the current published code; confirm every proposed sentence passes the adjudication test and the technology-neutral test; confirm your conflict-of-interest disclosure is drafted. Then route the outline twice: file it (or its relevant section) the next time your association opens any comment or guidance process, and attach it to your application the next time the state chapter ethics committee or national task force issues a call. Done looks like: a code-anchored, five-provision submission with vignettes and answered objections, a drafted disclosure statement, and two routing commitments on your calendar, the next comment window and the next volunteer call.

Key Takeaways

• An ethics-code revision is the constitutional convention of a profession: position papers, white papers, and conference panels are pamphlets that bind no one, while the code is what ethics committees, licensing boards, courts, and supervisors actually apply. The mission is the playbook's: serve on the standards-body committees that turn position papers into binding codes.
• The five codes have five cultures: APA's litigated, near-statutory enforceable standards; NASW's values architecture around 1.07; AAMFT's relational consent problems under 1.6 (multiple clients, one transcript); ACA's existing Section H technology framework, where the work is amendment; and the BACB's enforcement-forward 2022 code serving a Medicaid-heavy, 97151-97158 documentation practice base. Proposals must be drafted in the target code's own vocabulary and cite its existing provisions by number.
• Revision machinery is episodic, relational, and textual: cycles open years apart, committee seats are staffed from the association's service ecosystem, and specific provision language beats position arguments. Between cycles, ethics-committee advisory opinions are the case law that revisions later codify, so shaping the opinions pre-drafts the revision.
• The clinician-drafter's five-provision agenda: total responsibility for signed records, the clinical-determination boundary (no AI assessment, diagnosis, risk determination, or treatment decision; AI never scores the CSSRS or makes the Tarasoff or mandated-report call), consent obtained in the clinician's own voice with platform pre-consent insufficient, competence extended to AI oversight, and supervisee AI disclosure in the supervision agreement. Every provision attaches to an auditable artifact.
• Influence is earned on a service ladder, member presence, state and division committees, reviewer pools, national task forces, the revision committee, and it must be climbed before the cycle opens. The contributor who brings practitioner ground truth (verification-pass error logs, consent vignettes, supervision disclosures) becomes load-bearing; the contributor who brings opinions is one vote.
• The craft is position-to-provision conversion: identify the actor, make the obligation behavioral and observable, keep responsibility non-delegable, attach a trigger artifact, and test every sentence against the adjudication test, could an ethics committee determine from a complaint file whether it was violated? Write technology-neutral, draft the duty rather than the tool, respect the aspirational-versus-enforceable layers, and answer the access, innovation, and enforceability objections inside the submission.
• Integrity is the admission ticket: disclose vendor relationships and recuse where conflict norms require, because standards work runs on trust and the code belongs to the profession and the public, not to the drafter's practice model. The Submission Outline you build now is both your day-one comment filing and your committee-service application portfolio.