← All 95 filings

Nathan Sabich

IndustryConsultantFiled August 18, 2026281 wordsFDA-2026-N-7874-0015

What they argued

RecovryAI’s one-line reading of the filing.

Q1 only: proposes four-axis model adding human-oversight and traceability axes; no position on permitting autonomy.

Themes it raises

2 of the 21 themes in the docket, each with the passage we counted, verbatim.
What makes a function high riskFDA Q1, Q2, Q5
“It omits three dimensions that are critical for real-world assessment: (a) traceability of the output to source data, (b) the availability and effectiveness of downstream human oversight, and (c) the temporal context of deployment (real-time clinical decision vs. asynchronous review).”
Whether the user can judge the outputFDA Q3, Q4
“A GenAI function that generates a patient-facing summary with no clinician intermediary poses fundamentally different risk than one that produces a draft note for clinician review, even if both occupy the same position on the two-axis grid.”

FDA questions it names

Questions this filing names by number.

Q1 · The two-axis risk framework

Coded positions

Where a position was recorded question by question.
Q1Does a two-axis framework, AI device activity and the consequence of relying on an incorrect output, capture the dimensions of risk?
Keep it, but add or change elements

Across the five cross-cutting questions

RecovryAI’s reading of the whole filing. Silence is never counted as opposition.
Patient-facing autonomyShould FDA permit patient-facing AI to act with meaningful autonomy within a defined scope?
No position stated
Proportionate evidenceShould evidence requirements scale with clinical risk rather than a uniform high bar?
No position stated
Postmarket relianceCan strong postmarket monitoring justify accepting more premarket uncertainty?
No position stated
Competency evaluationCan a device be evaluated on competency benchmarks and clinical confirmation against clinicians?
No position stated
Change controlCan devices on third-party foundation models be maintained under pre-specified change control?
No position stated
Autonomy acceptedThe highest level this filing accepts
Low-consequence work: Not stated
High-consequence work: Not stated
Read and coded by RecovryAI readers, September 12, 2026. The source text and highlighted passages appear below. Read the filing on regulations.gov ↗

The comment as filed

Comment submitted on regulations.gov. Passages we counted are highlighted.

FDA Question 1: To start, the two-axis framework is a sound starting point but is insufficient for GenAI. It omits three dimensions that are critical for real-world assessment: (a) traceability of the output to source data, (b) the availability and effectiveness of downstream human oversight, and (c) the temporal context of deployment (real-time clinical decision vs. asynchronous review). A GenAI function that generates a patient-facing summary with no clinician intermediary poses fundamentally different risk than one that produces a draft note for clinician review, even if both occupy the same position on the two-axis grid.

An Architecture to Address This: I have created an architecture that explicitly structures risk mitigation through multiple dimensions beyond device activity and consequence. The HITL Validation Points across the lifecycle define distinct human oversight checkpoints, each with named roles and defined responsibilities. I utilize a transparency and labeling framework that requires documentation of intended use, intended user, known limitations, subgroup performance, model characteristics, output interpretation guidance, and update/version information. I created a multi-gate quality checkpoint system that provides progressive risk assessment at design, data integrity, model metrics, model acceptance, business review, and final live deployment, each with designated KOL reviewers.

Recommendation to FDA: Expand the framework to a four-axis model: (1) device activity (current Axis 1), (2) consequence of incorrect output (current Axis 2), (3) degree of human oversight available at the point of output delivery (ranging from fully autonomous to mandatory clinician review), and (4) traceability of output to verifiable source data (ranging from fully traceable to opaque/generative). This four-axis model would better capture the actual risk surface of GenAI-enabled devices and align with IMDRF’s lifecycle management framework, which emphasizes human oversight as distinct a governance dimension