Clarification request: Gemini API "clinical practice" restriction for a psychotherapy documentation tool

Gemini API
,
1

Hello,

I am a licensed psychotherapist in Mexico specializing in couples and family therapy. I use the paid tier of the Gemini API through a Google Cloud project with an active billing account through a Software I have coded. Before continuing to use it, and selling it to other therapists, I would like a written clarification on how your terms apply to my use case.

Relevant terms

  • Gemini API Additional Terms of Service (effective March 23, 2026), Use Restrictions: “You may not use the Services in clinical practice, to provide medical advice, or in any manner that is overseen by or requires clearance or approval from a medical device regulatory agency.”
  • Google Cloud Service Specific Terms, Generative AI Services, Healthcare Restrictions: no use “for clinical purposes (for clarity, non-clinical research, scheduling, or other administrative tasks is not restricted)”.

How the software works

  • It is a desktop application. All records are stored locally, encrypted, on the therapist’s own computer. No data is sent to me or to any server other than the Gemini API.
  • With each patient’s prior written informed consent, which explicitly covers AI processing, the transcript of a therapy session is sent to the Gemini API.
  • Gemini returns a documentation draft only: a session summary, the interventions the therapist made, progress and difficulties as discussed in the session, and the homework agreed on in session. It also drafts a short progress note for the clinical record.
  • The prompts explicitly instruct the model not to propose diagnoses, diagnostic hypotheses, treatment recommendations or prognoses. Those fields are left blank and are written by the therapist.
  • The therapist reviews, edits and approves every draft before it becomes part of the record. No decision is made automatically, and patients never interact with the model.

Questions

  1. Does the use described above (summarizing therapy sessions and drafting progress notes that a licensed therapist reviews and approves) fall under the “clinical practice” restriction, or is it considered permitted documentation or administrative use?
  2. Would it be permitted for the model to also generate diagnostic hypotheses or suggested therapeutic strategies, as non-binding input that the licensed therapist evaluates, be and may discard?
  3. If either use is not permitted under the Gemini API terms, agreement (for example, Vertex AI under a specific healthcareagreement) under which it would be permitted?
  4. Does removing names and other direct identifiers from transange the answer?
  5. If I license this software to other licensed therapists, each using their own Google Cloud project, billing account and API key, is there any
    additional restriction I should be aware of?
  6. Please confirm that the Data Processing Addendum for Products Where Google is a Data Processor applies to paid Gemini API usage without a separate
    acceptance step. Please also tell me for how long prompts a abuse monitoring.

I would appreciate a written response that I can keep on file.

Thank you,