Comment from EVRESA LLC

Please see the attached comments of EVRESA LLC regarding Docket No. FDA-2026-N-4390: Comments of EVRESA LLC — Docket No. FDA-2026-N-4390 EVRESA LLC, an AI-governance infrastructure company based in Orlando, Florida, respectfully submits these comments in response to the Request for Information on the AI-Enabled Optimization of Early-Phase Clinical Trials Pilot Program. Our comments address the trustworthiness, participant safety, data integrity, and governance infrastructure dimensions of the proposed pilot; we do not comment on the clinical optimization methodology. These comments are offered as policy input and do not include confidential or proprietary information. EVRESA's central submission is that the FDA should evaluate AI in early-phase trials not only by whether it improves the speed and quality of decisions, but by whether each AI-influenced decision is governed, human-authorized, privacy-minimized, and reconstructable. The accountability and explainability principles aligned with the NIST AI RMF are properties of the decision record and the governance process — not solely of the model — and can be enforced independently of whether the underlying AI is sponsor-developed or a proprietary third-party system that the FDA may not be able to inspect fully. We respectfully recommend that the pilot incorporate decision-level governance metrics, including: whether each consequential AI-influenced decision passed through a required human-authority gate; whether a complete, signed authority receipt exists and supports decision-provenance completeness; whether the receipt chain is tamper-evident and independently verifiable; the latency between AI recommendation, human review, and final disposition; and whether safety, drift, missing-evidence, or disagreement conditions triggered HOLD, escalation, or human override. Because these metrics are model-agnostic, they are directly comparable across sponsor-developed and proprietary systems alike. The attached comments respond to specific RFI questions on trustworthy AI and industry alignment, the roles of investigators and patient representatives in governance, shared governance infrastructure, accommodating different levels of AI maturity, pilot milestones, AI–human concordance and the interpretation of override rates, participant safety and data integrity, model-drift detection, safety and risk mitigation, trustworthiness metrics for proprietary systems, privacy and data governance, and subgroup fairness. Appendix A provides two illustrative, technology-neutral workflow figures. EVRESA's full comments, including Appendix A, are attached. We would welcome the opportunity to provide a technical briefing to the pilot program team. Respectfully submitted, John Gray Rodriguez Founder & Chief Executive Officer EVRESA LLC EVRESAai.COM