AI Ethics for Forensic Psychologists: What the Guidance Actually Says in 2026

Five documents. Two verified absences.

Most writing on AI ethics in psychology stays at the level of general principles. What a forensic evaluator needs is narrower: which documents carry formal authority, what each one actually requires, and where the guidance runs out. Here is the complete inventory, as of July 2026:

• 1. APA Ethical Guidance for AI (2025). “Ethical Guidance for AI in the Professional Practice of Health Service Psychology” — created June 2025, updated December 2025. Six considerations, none of them forensic-specific.
• 2. ABFP AI Policy (2026). The American Board of Forensic Psychology’s “Policy on the Use of Artificial Intelligence in Practice Samples,” approved April 18, 2026 — the newest formal AI guidance in the field and the centerpiece of this page.
• 3. Grabb & Angelotta (2023). The JAAPL editorial that widened the forensic AI conversation beyond machine-learning risk scores to large language models.
• 4. Roof (2025). The JAAPL commentary on AI-assisted deception — evaluees using LLMs to coach convincing symptom presentation.
• 5. Rilen (2026). A peer-reviewed article proposing an operational framework for integrating generative AI into forensic psychology practice.
• Absence 1: AAPL. The American Academy of Psychiatry and the Law has published no AI position statement, practice guideline, or resource document as of July 2026.
• Absence 2: AP-LS. The American Psychology-Law Society has issued no AI-specific guidance; the 2013 Specialty Guidelines for Forensic Psychology remain the operative document.

APA’s 2025 guidance: six considerations, zero forensic provisions.

APA’s “Ethical Guidance for AI in the Professional Practice of Health Service Psychology” (created June 2025, last updated December 2025) is the discipline’s baseline. It organizes its guidance under six considerations for using AI in practice:

• Transparency and informed consent
• Mitigating bias and promoting equity
• Data privacy and security
• Accuracy and misinformation risks
• Human oversight and professional judgment
• Liability and ethical responsibility

In substance: psychologists are expected to tell patients and relevant third parties when AI is used and obtain informed consent in culturally and linguistically appropriate ways; to vet tools for bias so AI does not widen health-care disparities; to use tools only in a HIPAA-compliant way with strong cybersecurity; to critically check AI output for accuracy before relying on it, and to stop using tools that produce misinformation. Throughout, AI is positioned as augmenting — never replacing — the psychologist’s judgment, with the psychologist retaining responsibility and liability for professional decisions.

What the document does not contain is any forensic-specific provision. Its only court-related mention is listing “the court” as an example of a third party that may be the client of psychological services for disclosure purposes. Nothing on expert testimony, admissibility, adversarial scrutiny of AI-assisted work, or the distinct informed-consent posture of a compelled forensic evaluation. For a forensic evaluator, the 2025 guidance sets the floor — it does not answer the questions specific to courtroom work.

The ABFP AI policy: the gap was named in 2024, then closed in 2026.

July 2024: the gap gets documented

In July 2024, four forensic psychologists — Hodges, Armstrong, Formon, and Silber — published “Implications for Artificial Intelligence in Forensic Psychological Practice and Board Certification” in ABPP’s newsletter On Board with Professional Psychology (Vol. 2, Issue 2, July 17, 2024). Among their observations: the ABFP Practice Sample Guidelines, updated as recently as May 2024, contained no guidance on AI assistance in submitted samples, and it was unclear how the sole-authorship and no-outside-review rules applied to AI. They also documented AI performing poorly on complex forensic reasoning, the tension between proprietary black-box algorithms and the Daubert testability and error-rate factors, and the need for experts to be transparent about AI limitations when testifying.

April 18, 2026: the policy that closed it

On April 18, 2026, the ABFP Board of Directors approved the “Policy on the Use of Artificial Intelligence in Practice Samples” — the newest formal AI guidance in forensic psychology and, as of July 2026, the only one written specifically for forensic work product. It defines AI broadly — content generation, summarizing, synthesizing, brainstorming, with examples including Grammarly, ChatGPT, and BastionGPT — while expressly excluding spell-check, ordinary web searches, and test-administration and scoring platforms such as Q-global and PARiConnect. Candidates must be able to do all of the work independently, and all practice samples may be screened with AI-detection software.

What it prohibits

• Using AI to administer, score, or interpret psychological testing.
• Using AI to draft or revise any portion of the work product or its conclusions and opinions — including accepting AI edits and presenting them as one’s own.
• Including AI-generated case-law, literature, or record-review content without independent verification.
• Relying on unverified AI transcription.
• Using any tool that is not HIPAA compliant.

What it permits

• Literature and case-law searches, where every result is independently verified.
• HIPAA-compliant AI record review, with confidentiality protections and personal verification of the output.
• Charts and visual aids.
• Administrative tasks such as consent forms and referral emails.
• Jurisdiction- or employer-mandated AI use, with documentation.

The disclosure requirement

If AI was used, the practice sample’s Introduction must disclose the tool, the rationale for using it, how it was used, the data-protection steps taken, and the ethical and legal considerations weighed. Supplemental materials must include the prompts, the AI outputs, and any training given to the tool. Noncompliance can mean rejection of the practice sample and termination of candidacy.

One precision note: this policy governs board-certification practice samples, not general forensic practice. But because it is the only formal document in the field that draws concrete permitted-versus-prohibited lines, its distinctions — verification before inclusion, HIPAA-compliant tooling, disclosure of tool, rationale, and data protection — are the closest thing forensic psychology currently has to a working standard, and a natural reference point for everyday practice.

What the forensic psychiatry literature adds.

Grabb & Angelotta (2023): the field widens its lens

Grabb and Angelotta’s editorial, “Emerging Forensic Implications of the Artificial Intelligence Revolution,” J. Am. Acad. Psychiatry Law 51(4):475–479 (2023), argued that forensic AI discussion had focused narrowly on machine-learning suicide-risk stratification, and that the arrival of consumer LLMs demanded a wider view. Their key points: there is almost no case law or legal precedent governing AI in medicine, leaving liability and admissibility unsettled; built-in LLM output variability, combined with missing citations and absent external validation, undermines reliability for clinical-legal use; models trained on biased datasets will reinforce that bias in forensic contexts; and chatbot tools raise data-encryption and HIPAA concerns. Their practical counsel: treat AI literacy as continuing education, and ask patients about their LLM use the way clinicians already ask about supplements.

Roof (2025): AI-assisted deception enters the picture

Roof’s commentary, “AI-Assisted Deception and the Emerging Challenge of LLMs in Forensic Psychiatry,” J. Am. Acad. Psychiatry Law 53(2):157–159 (2025), acknowledges the real gains generative AI offers forensic work — administrative tasks, report generation, record summarization — and then names the flip side: evaluees can use LLMs to coach convincing feigning of psychiatric symptoms. Alongside privacy and model-bias concerns, Roof flags fabricated or misleading AI-generated material entering forensic assessments, notes that legislation and professional guidelines for responsible use are still being built, and calls on forensic psychiatrists to shape responsible integration through education, advocacy, and best-practice development. For evaluators, the takeaway is double-sided: AI changes both how reports get scrutinized and how evaluees prepare for evaluations.

What does not exist — as of July 2026.

Knowing where the guidance runs out is as useful as knowing what it says. Each of the following absences was verified directly against the issuing organization’s publications:

• AAPL has no AI position statement. As of July 2026, the American Academy of Psychiatry and the Law has published no position statement, practice guideline, or resource document on artificial intelligence. Its AI engagement is confined to JAAPL journal articles.
• AP-LS has no AI-specific guidance. The American Psychology-Law Society’s operative document remains the 2013 Specialty Guidelines for Forensic Psychology, which predate modern generative AI entirely. AP-LS engagement so far is conference programming, not formal guidance.
• APA has nothing forensic-specific. Beyond the 2025 health-service-psychology guidance described above, APA’s only other formal AI document is the policy resolution “Artificial Intelligence and the Field of Psychology,” adopted by the Council of Representatives in August 2024 — discipline-wide, with no forensic mention.

The practical consequence: a forensic evaluator asked on the stand “what professional guidance governs your use of AI?” is working from a general APA document, one board-certification policy, journal commentary, and their own documented judgment. That makes individual disclosure and documentation practices carry more weight, not less.

A peer-reviewed integration framework.

The forensic-specific gap left by APA’s general guidance is exactly what Rilen’s “Practical Frameworks for Integrating Artificial Intelligence in Forensic Psychology” targets — published online February 21, 2026 in the Journal of Forensic Psychology Research and Practice (advance article, no volume or issue assigned yet). Its four-step framework, in one sentence: establish operational mandates and practitioner readiness, risk-stratify tool selection by distinguishing convergent from divergent tasks, separate substantive from peripheral AI uses to drive disclosure decisions, and analyze admissibility under Daubert and Frye. The article frames expert credibility as resting on methodological verification and procedural rigor rather than trust in the algorithm — a framing consistent with everything above, from APA’s human-oversight consideration to the ABFP policy’s verification requirements.

What the guidance suggests for everyday forensic practice.

Reading the five documents together, a consistent set of practices emerges. These are our synthesis — not requirements from any single document — framed the way the guidance itself frames them: as considerations.

• Disclosure. Consider adopting the ABFP policy’s disclosure structure for any AI-assisted work: name the tool, state the rationale, describe how it was used, document the data-protection steps, and note the ethical and legal considerations weighed. Courts have treated unverified, undisclosed AI use as the aggravating factor — see our review of the case law on AI disclosure.
• Informed consent. Consider addressing AI use in the notification-of-purpose conversation, consistent with APA’s transparency consideration — adapted to the forensic posture, where the retaining party and the court may be the relevant parties to inform.
• Tool selection. Consider asking of any tool: Is it HIPAA compliant, with a BAA? Where does the data go, and is it retained or used for training? Consumer tiers of general-purpose chatbots (ChatGPT Free and Plus, consumer Claude) operate without a BAA — OpenAI offers BAAs for its API and sales-managed enterprise arrangements, but not for its consumer products. Our analysis of the HIPAA exposure walks through the details.
• Verification. Consider treating every AI-surfaced citation, literature reference, and record summary as unverified until independently checked — the ABFP policy’s rule, and the lesson of every published AI-citation sanctions order. The hardest hallucination to catch is a real case cited for a holding it never reached.
• Documentation. Consider keeping the prompts and outputs for any substantive AI use, mirroring the ABFP supplemental-materials requirement — so that if AI use is ever raised in cross-examination, the record answers the question.

Structured review tools fit squarely inside the human-oversight consideration that runs through all of this guidance. ForensicShield, for example, reviews a completed forensic report for methodology vulnerabilities and admissibility exposure before the evaluator signs it — the evaluator retains full authorship, and the tool drafts nothing. It is HIPAA compliant with a signed BAA, encrypts report content with AES-256, runs AI inference inside a HIPAA-eligible AWS Bedrock boundary that PHI never leaves, passes every citation through a verification pipeline against public court databases with linked opinions and verification badges, and calibrates its analysis to 55 US jurisdictions. That is the division of labor the 2025–2026 guidance consistently describes: the psychologist is the expert; the tool assists with scrutiny, not authorship. See how it applies across evaluation types and jurisdictions, or run the manual version yourself with the Daubert self-audit checklist.

Common questions.

Is there APA ethics guidance specific to AI in forensic psychology?

No. As of July 2026, APA’s “Ethical Guidance for AI in the Professional Practice of Health Service Psychology” (June 2025, updated December 2025) applies to health service psychology generally and contains no forensic-specific provisions — its only court-related mention is “the court” as an example third party. The August 2024 Council resolution on AI is likewise discipline-wide.

What is the newest formal AI guidance for forensic psychologists?

The ABFP “Policy on the Use of Artificial Intelligence in Practice Samples,” approved by the ABFP Board of Directors on April 18, 2026. It governs board-certification practice samples: it prohibits AI drafting or revising work product or conclusions, requires independent verification of AI-surfaced case law and literature, requires HIPAA-compliant tools, and mandates disclosure of the tool, rationale, usage, and data-protection steps — with prompts and outputs attached as supplemental materials.

Do AAPL or AP-LS have AI position statements?

No. As of July 2026, AAPL has published no AI position statement, practice guideline, or resource document, and AP-LS has issued no AI-specific guidance — the 2013 Specialty Guidelines for Forensic Psychology remain the operative AP-LS document. AI engagement in both organizations is currently confined to journal commentary and conference programming.

Does the ABFP AI policy prohibit using AI altogether?

No. It draws explicit lines: prohibited uses include AI administering, scoring, or interpreting testing and AI drafting or revising any portion of the work product or its conclusions; permitted uses include independently verified literature and case-law searches, HIPAA-compliant record review with personal verification, charts and visual aids, and administrative tasks. The policy formally governs practice samples, but its permitted-versus-prohibited distinctions are a useful reference point for general forensic practice.

Every document, with links.

• American Psychological Association, Ethical Guidance for AI in the Professional Practice of Health Service Psychology (created June 2025; last updated December 2025) — apa.org.
• American Psychological Association, Artificial Intelligence and the Field of Psychology (policy resolution adopted by the Council of Representatives, August 2024) — apa.org.
• American Board of Forensic Psychology, Policy on the Use of Artificial Intelligence in Practice Samples (approved April 18, 2026) — policy PDF (listed in the ABFP document library).
• Hodges, H.J., Armstrong, N.E., Formon, D.L., & Silber, B.J., “Implications for Artificial Intelligence in Forensic Psychological Practice and Board Certification,” On Board with Professional Psychology, Vol. 2, Issue 2 (July 17, 2024) — abpp.org.
• Grabb, D.J. & Angelotta, C., “Emerging Forensic Implications of the Artificial Intelligence Revolution,” J. Am. Acad. Psychiatry Law 51(4):475–479 (2023), DOI 10.29158/JAAPL.230080-23 — jaapl.org (full-text PDF).
• Roof, J.G., “AI-Assisted Deception and the Emerging Challenge of LLMs in Forensic Psychiatry,” J. Am. Acad. Psychiatry Law 53(2):157–159 (2025), published online May 27, 2025, DOI 10.29158/JAAPL.250022-25 — jaapl.org (full-text PDF).
• Rilen, S., “Practical Frameworks for Integrating Artificial Intelligence in Forensic Psychology,” Journal of Forensic Psychology Research and Practice (advance online publication, February 21, 2026), DOI 10.1080/24732850.2026.2630383 — tandfonline.com.
• American Psychological Association, Specialty Guidelines for Forensic Psychology (2013) — apa.org.

Note on link availability: jaapl.org article pages were intermittently unavailable when this page was last verified; the full-text PDF links above are stable alternatives, as are the PubMed records (PMID 38065619 and PMID 40425270). This page is informational and does not constitute legal or ethical-compliance advice; guidance documents change, and the issuing organizations’ current publications control.