Psychological testing and cloud AI scribes: test security, interpretive feedback sessions, and the vendor archive that HIPAA alone cannot protect

Psychological testing in clinical and forensic practice — and where AI scribes enter the workflow

Psychological testing using standardized personality and clinical assessment instruments is a core service delivered by private practice psychologists in a range of contexts. The MMPI-3 (Minnesota Multiphasic Personality Inventory, Third Edition) is the most widely used objective personality measure in clinical and forensic practice in the United States. The PAI (Personality Assessment Inventory) is widely used in clinical settings, disability evaluations, and correctional psychology. The Rorschach — administered under the Comprehensive System or the more recent R-PAS scoring approach — is used in forensic contexts, complex diagnostic presentations, and treatment planning for personality pathology. The MCMI-IV (Millon Clinical Multiaxial Inventory) is commonly used in clinical practice focused on personality disorders and Axis II presentations.

The settings in which these instruments are used in private practice are broad: treatment planning for complex or refractory clients, forensic custody evaluations, disability determinations for SSDI and long-term disability insurance, pre-employment and fitness-for-duty evaluations, immigration psychological evaluations, presurgical psychological evaluations for bariatric surgery and SCS implantation, competency and criminal responsibility evaluations, and college counseling center assessments for academic accommodations.

In all of these settings, AI scribes are a natural efficiency tool. The clinical interview component of a psychological evaluation involves extensive structured and semi-structured questioning that an AI scribe can transcribe and organize into note content. Where AI scribes enter the workflow in assessment practice specifically, there are two distinct clinical sessions where the scribe may be present: the administration session (which includes the clinical interview, administration instructions, and often collection of the client's oral or written responses) and the interpretive feedback session (where the psychologist reviews and explains the test results with the client). Each creates a distinct and significant data custody problem when a cloud AI scribe is used.

The two-session structure of psychological testing

The administration session in psychological testing involves several components depending on the battery: a clinical interview covering the referral question, presenting concerns, personal history, and relevant background; standardized administration of the test instruments, which for the MMPI-3 and PAI involve the client completing written or computerized response forms; and for instruments like the Rorschach, oral administration with the client providing verbal responses to stimuli that the examiner records verbatim. The Rorschach protocol specifically requires the examiner to record the client's exact verbal responses during the association phase and the inquiry phase — responses that constitute test protocol content protected under APA Standard 9.11.

If a cloud AI scribe is active during a Rorschach administration, the vendor's session audio captures the client's verbatim responses to the Rorschach stimuli — the raw response protocol that psychologists are ethically required to protect from unqualified access. For MMPI-3 and PAI, the vendor's audio captures the administration instructions and any oral clarification the psychologist provides about specific items, as well as any verbal responses the client makes while completing the form. The vendor's archive includes this content alongside the clinical interview material.

The interpretive feedback session is a different clinical event and, from a data custody perspective, often the more consequential one. In professional practice, the interpretive feedback session is a structured 45-to-90-minute meeting in which the psychologist presents the test findings to the client, explains what the scale scores and profile patterns mean in accessible terms, and invites the client's response to the findings. This is not a brief review of results — it is a full clinical session, and it generates a specific kind of disclosure that rarely appears in any other clinical encounter: the client's direct, immediate reaction to hearing a structured psychological description of themselves.

What psychologists who conduct feedback sessions consistently observe is that the feedback session often produces the client's most significant and least guarded disclosures. A client who has been relatively controlled throughout the clinical interview may, upon hearing that their MMPI-3 profile shows a marked Scale 6 elevation, spontaneously explain a paranoid episode from their past they had not previously mentioned. A client with a PAI elevation on the Borderline Features scale may react to hearing the clinical description by sharing specific relational patterns they had minimized throughout the intake. A client with notable validity indicator elevations — the F scales, the RBS — may respond to the psychologist's characterization of their response style by explaining in detail the circumstances that led them to answer the items the way they did.

None of this content appears in the formal psychological test report. The formal report translates the raw profile into professional clinical language and frames the findings around the referral question. The feedback session is where the client and clinician negotiate that translation — and that negotiation, in all its verbatim texture, is what the cloud AI scribe vendor retains.

APA Ethical Principles Standard 9.11: test security and the vendor's unqualified access

APA Ethical Principles of Psychologists and Code of Conduct Standard 9.11 states that psychologists make reasonable efforts to maintain the integrity and security of tests and other assessment techniques consistent with law and contractual obligations, and in a manner that permits adherence to the Ethics Code. The Standard specifically includes protecting test materials from access by unqualified persons. Test materials are defined to include manuals, instruments, protocols, and test questions or stimuli — in short, the test content and the data generated during administration.

The Standard's primary concern is preventing compromise of test validity: if raw test items and scoring methods become widely available to non-psychologists, the normative integrity of standardized instruments is undermined, and the interpretive validity of future administrations is compromised. This is a profession-wide interest in maintaining assessment as a protected professional function.

The application to cloud AI scribes is direct. A cloud AI scribe vendor processes session audio through automated transcription systems and, in most implementations, through language models that generate structured notes from the transcript. The vendor's engineering team, data science staff, and security personnel may have legitimate access to session audio as part of normal operations — access that is authorized under the vendor's own data access policies and consistent with the BAA's minimum necessary framework for providing the contracted service. These individuals are not licensed psychologists. A business associate agreement governs the vendor's data security obligations, but it does not transform the vendor's staff into qualified persons under APA Standard 9.11's meaning.

When a psychologist uses a cloud AI scribe during a Rorschach administration — or during an MMPI-3 administration session that includes discussion of specific items — and the vendor's session audio includes the client's verbal responses to test stimuli, the psychologist has permitted test protocol content to reach unqualified persons. This is not a hypothetical APA Ethics Committee concern — it is a structural consequence of the cloud AI scribe architecture. The BAA does not cure this because the problem is not about data security; it is about access by unqualified persons to test materials.

For MMPI-3 and PAI testing, the exposure is somewhat lower at the administration stage because the client completes the test form independently rather than giving oral responses. But the feedback session, where the psychologist reads aloud and explains the scale names, item content, and scoring interpretations while the client's audio is being recorded by a cloud AI scribe, creates vendor access to a detailed explication of the scoring methodology, the clinical interpretation of scale elevations, and the client's spontaneous verbal responses to that explication — all of which qualify as test materials and test protocol content in the relevant APA sense.

The formal test report versus the interpretive feedback session content

The formal psychological test report is a professional document. For an MMPI-3 or PAI evaluation, a well-constructed report presents the validity scale analysis, the clinical scale profile, the specific scale elevations and their clinical significance, the diagnostic formulation supported by the testing, and the treatment or referral recommendations appropriate to the referral question. A forensic report for a custody evaluation presents the findings in relation to the specific parenting capacity question. A disability evaluation report presents the findings in relation to functional capacity and the diagnostic criteria for the claimed conditions.

What the formal report does not contain is the client's verbatim reaction to hearing their own profile. The feedback session that produced the final report reflects the client's response to the findings — psychologists routinely incorporate client feedback into their final report, noting when a client's reaction to a finding was consistent with or contradicted the interpretation — but what appears in the report is the psychologist's synthesis, not the verbatim session content.

When a client with a high PAI NIM (Negative Impression Management) score hears that finding and responds by explaining in detail the medical history they believe justifies their symptom reporting, that explanation is in the vendor's feedback session audio. It is not in the formal report, which notes the NIM elevation and the psychologist's conclusion about its significance. When a client with a notable MMPI-3 RC6 (Ideas of Persecution) elevation reacts to hearing that finding by disclosing a specific workplace incident they had not mentioned during the clinical interview, that disclosure is in the vendor's audio. When a client in a custody evaluation reacts to hearing a finding about their parenting stress by making statements about the other parent that they had previously withheld, those statements are in the vendor's audio.

In each case, the gap between the formal report and the feedback session content is the specific target in adversarial proceedings where the evaluation itself is contested.

Legal contexts where vendor-retained feedback session audio reaches adverse parties

Contested custody evaluations. Psychological testing is a standard component of forensic custody evaluations, with MMPI-3 and PAI routinely administered to both parents and in some cases to older children. In contested forensic evaluations, the evaluator's full methodology is subject to challenge. A parent whose custody evaluation was unfavorable can challenge the evaluation on methodological grounds, including the adequacy of the feedback session. A third-party subpoena directed at the cloud AI scribe vendor for the feedback session audio provides opposing counsel with the verbatim record of how each parent responded to their own psychological profile — content that the formal report does not contain but that may be highly probative when the parent claims the evaluator mischaracterized their responses or failed to incorporate their explanations for scale elevations.

Disability insurance proceedings. MMPI validity indicators — the VRIN-r, TRIN-r, F-r, Fp-r, and RBS scales — are the primary objective measures used to assess potential response exaggeration and symptom magnification in disability claimants. In SSDI and long-term disability adjudications, insurers and the Social Security Administration's medical review process specifically examine validity indicator results when functional limitations are contested. Disability proceedings generate extensive record requests and, when records are insufficient, third-party discovery. The feedback session audio in a disability-context evaluation may contain the psychologist's verbal explanation of the validity findings — and, critically, the claimant's verbal response to hearing that their response style was flagged. Insurance defense counsel seeking to characterize a claim as involving response exaggeration will attempt to obtain the feedback session audio for exactly this content.

ADA employment discrimination proceedings. Psychological testing in fitness-for-duty evaluations and return-to-work clearances is a specific context where Title I ADA employment discrimination claims generate litigation. When an employer orders a fitness-for-duty psychological evaluation and the employee subsequently challenges an adverse employment action — termination, demotion, failure to provide reasonable accommodation — the employee's attorney seeks all records related to the evaluation. The feedback session, where the psychologist discussed the findings with the employee, is particularly significant because it may contain the psychologist's characterization of the employee's mental health condition and functional limitations in more direct terms than the formal report uses for the employer's audience.

Criminal proceedings and forensic examinations. Psychological testing is a central component of competency evaluations, sanity assessments, and capital sentencing evaluations. Rule 35 court-ordered mental examinations involve formal psychological testing in adversarial contexts. In capital sentencing mitigation proceedings, the PCL-R (Hare Psychopathy Checklist-Revised) and MMPI data are key evidence. In any contested forensic evaluation in a criminal proceeding, both prosecution and defense have motivation to obtain the full record of the evaluation — including the feedback session audio that documents how the evaluating psychologist explained their findings and how the defendant responded. Subpoenas directed at third-party vendors in criminal proceedings have broad reach, and the vendor's record of the forensic evaluation feedback session is directly probative in challenges to the evaluator's methodology and the reliability of their conclusions.

On-device processing: the only architecture that satisfies both HIPAA and Standard 9.11

The data custody problem in psychological testing is a dual one. HIPAA requires that the vendor manage the client's health information under a BAA with appropriate security controls and minimum necessary access limitations. APA Standard 9.11 requires that the psychologist take reasonable steps to prevent access to test materials and protocol responses by unqualified persons — a requirement that a BAA does not satisfy because vendor personnel with legitimate operational access are unqualified persons for this purpose.

On-device processing eliminates both problems simultaneously. If the session audio from the administration session, the clinical interview, and the interpretive feedback session is processed on the psychologist's Mac and never transmitted to vendor infrastructure, the vendor holds no session content. There is no vendor archive of test protocol responses for Standard 9.11 purposes — the unqualified persons at the vendor never had access. There is no vendor archive of feedback session audio for HIPAA purposes — the vendor holds no PHI from the testing sessions. A third-party subpoena directed at the vendor produces nothing because there is nothing to produce.

The psychologist's own records — the formal test report, the structured notes from the clinical interview, the testing protocols retained in the client file — remain subject to the same subpoena and licensing board process they always were. On-device processing does not eliminate the psychologist's own documentation from legal process. What it eliminates is the independently custodied vendor archive: the second copy of the most sensitive session content, held by a party with its own legal obligations to respond to process, independent of the psychologist's professional framework.

For psychologists who conduct psychological testing as a meaningful part of their clinical or forensic practice, this architectural distinction is the central data governance question raised by cloud AI scribes. The formal test report is a professional document designed to be shared in appropriate clinical and legal contexts. The test administration session and the interpretive feedback session are clinical events that generate content the formal report distills but does not reproduce. That content — the client's verbatim responses to hearing their own psychological profile, their explanations for their scale elevations, their emotional reactions to the findings — belongs in the psychologist's own records under the psychologist's professional framework. As with neuropsychological evaluation documentation, the specific harm arises not from the formal record's disclosure but from the existence of a vendor-held verbatim record that the formal documentation framework was never designed to govern.

Frequently asked questions