Before you ever meet your first client on an EAP platform, someone has already consented for you. The member clicked through the platform's terms of service during enrollment, and somewhere in that flow, between the privacy policy and the appointment scheduler, sat a clause consenting to ambient session capture and AI-assisted documentation. The platform's master services agreement with the employer-purchaser promises outcomes data; the contractor agreement you signed, probably around §8, obligates you to use the platform's scribe stack. And your ethics code, the one your license answers to, says informed consent is something you obtain, in your voice, in the room. This lesson maps that collision at Lyra, Spring Health, Modern Health, Talkspace, BetterHelp, and Tava Health, and builds your counter-voice: a session-one consent script delivered no matter what the platform collected, a documented position that platform consent does not substitute for yours, and the hard limit at the end of the road.

The Officiant Who Was Handed a Signed Certificate

Hold this analogy through the lesson: the platform has handed you a marriage certificate signed before the wedding. Imagine an officiant arriving to perform a ceremony and being told the couple already signed the certificate at the venue's front desk, between the parking validation and the coat check, so the vows are optional, skip to the reception. Everything legally necessary may exist on paper. Yet every officiant knows the vows are not decoration. The vows are the event: two people saying out loud what they are agreeing to, with the person who solemnizes it as witness. Consent in psychotherapy is the same kind of thing. The platform's click-through is the certificate from the front desk; your session-one conversation is the vows. Your ethics code is unambiguous about which one it requires.

The collision is structural, not accidental. An EAP platform sells to employers, and the product is increasingly not just access to therapy but measurable outcomes: engagement rates, symptom trajectories, return-to-work data. Outcomes at that scale require structured documentation at that scale, which is why the platforms moved early on AI scribes. Standardized capture across thousands of contracted clinicians only works if consent is standardized too, so consent moved upstream: into the member terms of service, collected once, at enrollment, before any clinician is assigned. From the platform's perspective this is elegant. From your ethics code's perspective, it is a category error: informed consent treated as a data-processing permission when your profession defines it as a clinical process.

Meet the clinician in the intersection. Dana is an LCSW in Sacramento, three years on a major EAP platform's network, twenty-two platform clients a week. Last month the platform activated ambient capture for her sessions. No conversation preceded it; an email announced "documentation enhancements," and the clinician portal confirmed that members "have already consented via the Terms of Service." Dana's first thought was relief, eight fewer notes a night. Her second thought, the one that kept her up, is the question this lesson answers: if the client never heard about recording from me, in my voice, have I met my ethics-code obligation, and what happens to my license if the answer is no?

The Three Contracts and the One License

Lay out the documents in the collision; each binds someone, and only one binds you to your board. Document one: the platform-employer MSA. The master services agreement between platform and employer-purchaser is the commercial engine. It typically includes outcomes-data sharing commitments (aggregate engagement and symptom-trajectory reporting to the employer) and, increasingly, AI-scribe deployment as the documentation infrastructure producing those outcomes. You are not a party to it and have probably never seen it. But it is why the scribe is in your sessions.

Document two: the member-facing terms of service. When the employee enrolls, the ToS flow collects agreement to the platform's data practices, and at the platforms this lesson names, that flow now includes ambient-capture and AI-documentation consent. This is the pre-consent: gathered before the member has met you, bundled with scheduling permissions and notification preferences, in the platform's voice, serving the platform's purposes.

Document three: your contractor MSA. The agreement you signed to join the network. At Lyra, Spring Health, Modern Health, Talkspace, BetterHelp, and Tava Health, the AI-scribe provisions tend to live in the same structural neighborhood, around §8 of the contractor MSA, the section governing documentation, technology use, and data. Read your own §8 with these questions: Does it obligate you to use the platform's scribe stack, or merely permit it? Does it state that the platform handles member consent, and does it address your right to conduct your own consent conversation? Does it restrict you from telling the member how their session data is processed? Does it address a member who, having heard your explanation, wants out? Most clinicians signed this document the week they were credentialed and have not read §8 since.

Then there is the document that outranks all three for you: your ethics code. APA Standard 4 (privacy and confidentiality, with informed consent for recording), NASW 1.07 (including consent before recording), AAMFT 1.6, and ACA Section H converge on one requirement: the clinician obtains meaningful informed consent, as part of the clinical relationship, in the clinician's voice, not the platform's. Meaningful is the operative word. A clause on page eleven of an enrollment flow, agreed to before the client knew your name, is consent to the platform's data processing. It is not the clinical informed consent your code requires, because the client never had the conversation in which the clinician they are entrusting with their inner life explains what is captured, what it becomes, who can see it, and what saying no would mean.

Why the Boards Side with the Code, Not the Contract

If the conflict reaches your licensing board, the platform's paperwork will not save you. The CA BBS, NY OPP (the Office of the Professions, governing licensed mental health practitioners in New York), and TX BHEC (the Texas Behavioral Health Executive Council) have all addressed contractor-platform ethics primacy in recent guidance, and the through-line is consistent: a licensee's professional obligations are not delegable to, or dischargeable by, a corporate platform's terms of service. The contractor relationship creates no exemption from the consent, confidentiality, and documentation standards that attach to the license. In board logic, the platform is a setting, not a shield. You can practice in someone else's building, on someone else's referral stream, under someone else's technology stack, and the board still examines your conduct as if you chose every element, because you chose one: to practice there.

Run the complaint scenario. A member discovers, eight sessions in, that their sessions have been processed by an AI scribe. They do not remember consenting; they never discussed it with their therapist. They file with the state board. The board's questions go to the licensee, not the platform: Did you know the sessions were being captured? (Yes.) Did you inform the client and obtain consent? (The platform did, via the ToS.) Show us where in your clinical record the informed-consent discussion is documented. (There is no entry.) At that point, the beautifully drafted ToS is evidence in the platform's defense, not yours. The platform's lawyers do not represent you. Its consent does not appear in your chart. And the board is applying NASW 1.07 or its state equivalent to you, the human with the license, who sat in the room for eight sessions knowing something the client did not.

The platform's consent lives in its database; your consent obligation lives in your license. A board complaint will be addressed to the second one.

The Counter-Voice: Your Session-One Script

The resolution is not to fight the ToS; it is to render it clinically irrelevant by doing your job anyway. Deliver your own consent script at session one, in your voice, regardless of what the platform collected, and document that you did. Five movements; here is a working version Dana could deliver:

"Before we start, I want to tell you how I document our work, because you deserve to hear it from me and not from a terms-of-service page. This platform uses a tool that listens to our sessions and drafts my clinical notes; I review every word and I sign them, so the note in your record is mine, not the software's. The audio and draft are processed under privacy agreements, and I can tell you more about where that data goes if you want the detail. You may have agreed to something like this when you enrolled, but what matters to me is whether you agree now, after hearing it from me. If you would rather not have the tool running, tell me, today or anytime, and I will do everything in my power to turn it off for your sessions; saying no will not change your care or our relationship. Do you have questions, and are you comfortable proceeding?"

Notice what the script does. It re-locates consent in the clinical relationship (you hear this from me). It tells the truth about the mechanism and the signature (the tool drafts, I review and sign, the note is mine). It offers depth without forcing it. It subordinates the pre-consent to the present conversation. And it creates a real, ongoing opt-out with a no-penalty promise. Then the documentation, a chart entry for every new client: "Discussed AI-assisted documentation with client in session one, including capture mechanism, clinician review and signature, data handling, and the client's right to decline now or at any time without effect on care. Client consented verbally. Platform ToS consent noted as preceding but not substituting for this discussion." That last sentence is your position statement in miniature, entered contemporaneously, in the record the board will read.

The mid-episode opt-out deserves its own mechanics, because a right the client cannot exercise is a flourish, not a safeguard. Before relying on the script, verify with platform support, in writing, exactly how capture is disabled for a single client, how long it takes, and how you confirm it happened. If the platform has no per-client off switch, you need to know before you promise one, and the absence is itself a finding for this lesson's last section.

The Position Statement in the Supervision File

The second artifact is the documented position: a short memo, dated and kept where your professional record lives, stating that platform consent does not substitute for clinician-administered informed consent. If you are supervised, or supervise others on platform contracts, it goes in the supervision file; if independent, in the compliance file next to the malpractice policy. The memo does three jobs. It evidences contemporaneous professional judgment: if a complaint arrives, you can show you identified the consent gap and closed it as documented practice, not post-hoc rationalization. It protects supervisees: a supervisor whose associates work platform caseloads should require this position and the script in the supervision agreement, because the associate's consent failure is the supervisor's board exposure. And it disciplines you: a signed position is one you will notice yourself violating on the tired Tuesday when skipping the script would be easier.

The memo's core paragraph can be this plain: "It is my documented position that consent collected by [platform] through its member terms of service constitutes the platform's data-processing authorization and does not substitute for the clinician-administered informed consent required by [NASW 1.07 / APA Standard 4 / AAMFT 1.6 / ACA Section H, as applicable]. I deliver my own consent discussion at the first session with every platform client, document it in the clinical record, and maintain a per-client opt-out. Where the platform's systems or contracts prevent me from fulfilling this obligation, I treat that as an ethics conflict to be resolved in favor of the code." Date, license number, signature. Ten minutes of writing; years of cover.

One guardrail belongs here explicitly: AI cannot resolve this conflict for you. No prompt, no policy generator, no contract-summarization tool can make a platform agreement compatible with an ethics code when the two actually conflict. AI can draft the script, format the memo, and organize your §8 review, all downstream of your judgment. The judgment itself, whether your consent obligation is met on this platform, with this client, under this contract, is a licensed human's call, exactly the way a risk determination is. A clinician who delegates an ethics-code conflict to a language model has misunderstood both the model and the code.

Reading Your §8: The Contract Audit

Now the audit. Pull your contractor MSA for every platform you work with, find the documentation-and-technology section (start around §8), and answer six questions in writing. One: is scribe use mandatory, default-on, or elective? Two: does the contract assert that member consent is handled by the platform, and does it anywhere restrict your own consent conversation? Three: is there a per-client opt-out, and who can trigger it: you, the member, or only the platform? Four: what does the contract say about session audio and derived data: retention, model training, de-identification claims, the subprocessor chain? Five: what are you indemnifying the platform for, and is a consent-related board complaint inside or outside that indemnity? Six: what are your termination rights and what happens to clients in your care if you exit?

The answers sort platforms into three postures. Compatible: the contract permits (or does not prohibit) your own consent conversation, a per-client opt-out exists and works, and nothing in §8 forces capture over a client's refusal. Here the script plus memo plus documented opt-out mechanics bring you into alignment; practice in peace. Negotiable: the contract is silent or ambiguous, the opt-out is unclear, or capture is default-on with no stated refusal path. Escalate in writing: ask the platform's clinical leadership the six questions, request written confirmation of the opt-out mechanics, and put your position memo on the record. Platforms answer to credentialing bodies and enterprise clients; a clinician asking precise, documented consent questions is hard to ignore, and clinicians asking together, through professional associations, are harder still. Incompatible: the contract or systems prevent you from delivering your own consent script, or force capture over a client's refusal, and escalation has not fixed it.

The Hard Limit: When You Leave

State the limit without softening it, because every escalation path needs a floor: a clinician who cannot deliver their own consent script on a platform should leave the platform. Not "should feel conflicted," not "should raise at the next town hall": should leave, on a planned timeline that protects clients currently in care. The reasoning is the spine of this program. Your license is the asset that matters: the thing that took years to earn, that feeds your family, that every client's trust silently relies on, and the only thing in this collision that cannot be re-issued by a sales team. The contract is replaceable; the caseload is rebuildable; the license is neither. A platform whose §8 forces you below your ethics code is asking you to spend license to make its unit economics work, and no per-session rate prices that correctly.

Leaving well is its own small protocol. Review the termination clause and notice period. Plan continuity for active clients: completion of current episodes where the contract allows, warm referrals where it does not, the transition documented in each chart. Put your reason in writing to the platform, citing the specific consent incompatibility: a paper trail of clinicians leaving over consent is the signal that eventually changes architecture. And tell your professional association: the APA, NASW, AAMFT, and ACA ethics infrastructure moves on member reports, and the contractor-platform consent question is exactly the kind of issue the recent CA BBS, NY OPP, and TX BHEC guidance grew out of. The individual exit protects your license. The documented, reported exit helps move the industry. Both are your job at this level.

The Applied Problem: Build the Session-One Script and the Position Statement

Your artifact is a two-part package: the Session-One Consent Script and the signed "Platform Consent Does Not Substitute" Position Statement, built against a real contract.

Step one: run the §8 audit. Pull your contractor MSA (or request it before credentialing) and answer the six audit questions in writing: mandatory versus elective scribe use, platform consent assertions and restrictions on your own conversation, the per-client opt-out and who controls it, data retention and training and subprocessors, indemnity scope, and termination mechanics. Classify the platform: compatible, negotiable, or incompatible.

Step two: draft your script in your own clinical voice, keeping the five movements: you hear this from me; the tool drafts and I review and sign; data detail on request; your agreement now is what matters, not the enrollment click; a real opt-out, now or anytime, with no effect on care. Read it aloud once; rewrite anything you would not say to a person. Verify the opt-out mechanics in writing before the script promises them.

Step three: write and sign the Position Statement using the core paragraph from this lesson, citing your ethics code by section, and file it: supervision file if supervised or supervising, compliance file beside the malpractice policy if independent. Pair it with the standing chart-entry language for every session-one consent discussion, including the sentence noting that platform ToS consent preceded but did not substitute.

Step four: the verification pass. Role-play the script with a colleague playing a client who asks, "Wait, I never agreed to recording," and a second pass where the client says, "I agreed at signup but I want it off now." Your answers must be true under your platform's actual mechanics; if not, your platform is negotiable or incompatible, and your next artifact is the written escalation or the exit plan. Done looks like: a completed six-question audit with a posture classification, a script deliverable from memory in under ninety seconds, a signed and dated position statement in the right file, the chart-entry language saved as a template, and written confirmation of the opt-out mechanics, or a documented escalation in its place.

Key Takeaways

• EAP platforms pre-consent members to ambient capture and AI documentation through the enrollment terms of service, before the member meets the clinician, because the platform-employer MSA sells outcomes data requiring standardized capture. The click-through is the certificate at the front desk; your session-one conversation is the vows.
• Three documents collide: the platform-employer MSA (outcomes-data sharing and scribe deployment), the member ToS (the pre-consent), and your contractor MSA, whose AI-scribe provisions typically sit around §8 at the six named platforms. Only your ethics code binds your license: APA Standard 4, NASW 1.07, AAMFT 1.6, and ACA Section H all require meaningful consent in the clinician's voice.
• Boards side with the code: CA BBS, NY OPP, and TX BHEC guidance converges on contractor-platform ethics primacy. The platform is a setting, not a shield; in a complaint, the board's questions go to the licensee, and the ToS is the platform's defense, not yours.
• The counter-voice has two artifacts: a five-movement session-one script delivered regardless of what the platform collected (you hear it from me; the tool drafts and I sign; detail on request; your agreement now is what matters; real opt-out anytime without penalty), and a signed position statement that platform consent does not substitute, filed with matching chart-entry language.
• Verify the opt-out before promising it: confirm in writing how per-client capture is disabled. A right the client cannot exercise is a flourish, and a missing off switch reclassifies the platform.
• AI never resolves an ethics-code conflict with a platform contract. It can draft the script and format the memo downstream of your judgment, but whether your consent obligation is met is a licensed human's call, exactly like a risk determination.
• The hard limit: a clinician who cannot deliver their own consent script on the platform should leave the platform, on a planned timeline protecting active clients, with the reason documented to the platform and reported to the professional association. The contract and caseload are replaceable; the license is the asset that matters.