From the Research Universities edition of September 27, 2026
ORI Aug 2026 generative AI guidance: disclose AI use, verify results; detection tools not sole misconduct evidence
U.S. Department of Health and Human Services Office of Research Integrity (ORI) · Federal research integrity / PHS-funded research administration · United States (PHS-funded institutions nationwide)
- Publisher
- ORI Guidance on Generative Artificial Intelligence (42 CFR Part 93)
- Original publication
- August 2026 (issuance); ORI release page Aug 14, 2026
- Source retrieved
- 2026-09-28
What happened
ORI’s August 2026 nonbinding guidance explains how generative AI interacts with PHS research misconduct rules under 42 CFR Part 93 (2024). Researchers should identify generative AI tools used in research and in manuscript/grant preparation and verify results, including AI-derived outputs. Disclosure is not a safe harbor if use significantly departs from accepted practices of the relevant research community. Institutions should gather all available evidence and are advised against relying on AI detection tools as the sole source of evidence; plagiarism detectors may miss AI-generated scientific text. Institutional records and sequestered evidence are generally retained seven years. Why now: same-week physical instrument agents and embodied-lab benchmarks make disclose/verify/preserve controls operationally binding for research administration.
Why it matters
Research administration and RIOs adopting AI-assisted discovery need federal misconduct-aligned disclosure, verification, and record-retention practices that transfer across PHS-funded campuses.
Evidence and measured results
ORI PDF guidance opened via HTTPS; disclose/verify recommendations, non-sole-reliance on detection tools, and Part 93 framing taken from opened PDF (Date of Issuance August 2026; release page Aug 14).
Limitations and uncertainty
Nonbinding guidance; August date older than last edition—included for why-now control-layer relevance; does not evaluate scientific agent performance; no published compliance KPIs.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-09-27; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- PHS-funded units adopt generative AI and research agents faster than misconduct-aligned disclosure, verification, and record practices.
- Stakeholders
- Research Integrity Officer / ORIA
- VPR / sponsored research
- Compliance / counsel
- PIs and lab managers
- Library / research data services
- Discovery questions
- Where must generative AI use be disclosed in grants and manuscripts?
- How are AI prompts/outputs retained as research records?
- Are AI detectors used as sole misconduct evidence?
- Do institutional AI policies define accepted practices for committees?
- Value hypothesis
- Aligning campus AI research workflows to ORI disclose/verify/preserve guidance can reduce uncontrolled generative-AI use risk in PHS-funded settings.
- Potential engagement
- Gap assessment of disclosure templates, AI artifact retention, and inquiry SOPs against ORI generative-AI guidance.
- Unsupported claims
- Measured drop in misconduct findings
- Safe harbor from disclosure alone
- Binding mandate for standalone AI policies
- Scientific quality improvement from policy alone
Pre-sales engineering
Role takeaway
- Fit
- Research platforms and agent gateways that capture AI tool identity, prompts/parameters/versions, and human verification attestations in the research record.
- Architecture and integration
- Connect grant/manuscript systems, LIMS/ELN, and agent logs to sequesterable storage with RIO access paths under Part 93 timelines.
- Prerequisites
- Research-record definition covering AI artifacts
- Disclosure fields
- Retention schedule (≥7 years when applicable)
- RIO access procedures
- Training
- Constraints
- Nonbinding guidance
- Community-practice dependence
- August issuance age
- No KPI dashboard in the guidance
- Security
- Protect sequestered AI evidence; control access during inquiry/investigation; coordinate with IRB/IACUC retention.
- Proof of value
- Disclosure appearing in sample outputs
- AI artifact retention tests
- Inquiry SOP forbids detector-only findings
- Training completion for PIs/RIOs
Delivery
Role takeaway
- Work
- Publish disclosure templates, update retention SOPs, train RIOs/PIs, revise inquiry checklists.
- Dependencies
- Sponsored research systems
- ELN/LIMS vendors
- Counsel
- College research offices
- Ownership
- RIO owns misconduct process; VPR owns research policy; units own local disclosure compliance.
- Skills and adoption
- Train PIs to disclose and verify; train committees on AI subject-matter consultation and evidence breadth.
- Governance checkpoints
- Policy/guidance crosswalk
- Retention audit
- Sample inquiry tabletop
- Acceptance criteria
- Disclosure templates live
- Retention path tested
- Detector-only ban in SOP
- Risks
- Disclosure without verification
- Shadow AI tools
- Incomplete prompt logs
- Guidance misread as optional forever
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Prompt/version/output retention and research-record systems must capture generative AI use trails usable in inquiry/investigation sequestration.
Governance
Who approves, reviews and stays accountable for outcomes?
RIO-led proceedings under Part 93; optional institutional generative-AI policies may help define accepted practices; funder AI policies may inform community norms.
Security and privacy
What data, permissions and controls need testing?
Evidence sequestration and seven-year retention apply to AI session artifacts when they are research records; coordinate with IRB/IACUC retention policies.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
RIO Boot Camp and AskORI technical assistance are the federal workforce supports named; campus training remains the local adoption layer.
Procurement
What should contracts, pricing and exit terms secure?
Prefer research tools that export sequesterable AI-use logs; do not procure AI detectors as sole misconduct adjudication systems.
Operating model
Which teams own the service once it runs?
Disclose–verify–preserve under Part 93 is the operating model for generative AI in PHS-funded research integrity.
Publication history
- 2026-09-27Research Universities · Issue 214 resources
Stable resource ID: ori-generative-ai-research-misconduct-guidance-202608