FDA Q-Sub strategy for AI agent governance.
How to use the FDA Pre-Submission program to lock in regulatory classification before deploying clinical AI agents. Written for Chief Medical Information Officers, Heads of Clinical AI and digital health founders preparing for an FDA conversation in 2026 or 2027.
Written by Ashish K. Saxena · Founder, Caventia
The cheapest way to lock in classification before you ship.
The single most consequential FDA decision for a clinical AI agent platform is not the 510(k) clearance. It is the classification call: device or non-device. Get it wrong and you are spending eighteen months and seven figures on a regulatory pathway you did not need to enter. Get it right and you can ship inside the 21st Century Cures Act Section 3060 carve-out for clinical decision support software.
The FDA Pre-Submission Program (Q-Sub) is the mechanism for locking in that call. It is free, voluntary and returns written FDA feedback within seventy to seventy-five days. For an AI agent governance platform that audits, logs and governs other AI systems without making clinical decisions itself, the Q-Sub is the gate that confirms non-device status before customer contracts make refactoring expensive.
The framing has to be precise. The platform does not acquire, process or analyze patient data. It captures metadata about the operation of customer-owned AI systems. The platform does not generate clinical recommendations. Recommendations are produced by customer systems. The platform supports the independent clinical review required under Section 3060. Each of those sentences corresponds to one of the four Section 3060 criteria that determine whether you are inside the carve-out or outside it.
The whitepaper covers the package contents, the five specific questions you ask FDA, counsel selection criteria, risks and mitigations and the six-month timeline that lets a Q-Sub return before your Healthcare SKU launch. It also covers what you cannot say in marketing without triggering FTC unfair-trade-practice exposure (FDA approved, FDA cleared, FDA certified are all off the table for governance software).
If you are a clinical AI founder reading this in 2026, filing a Q-Sub before Healthcare launch is the highest-ROI regulatory move you can make this quarter. [Download to continue reading.]
Nine sections and an appendix on Healthcare regulatory hooks.
Written for clinical AI founders and Chief Medical Information Officers approaching FDA for the first time.
I.
What a Q-Sub is and is not
The FDA Pre-Submission Program is a free, voluntary mechanism for written FDA feedback within 70 to 75 days. It is not a clearance. It is the cheapest way to lock in classification before customer contracts make refactoring expensive.
II.
The Section 3060 carve-out
Under the 21st Century Cures Act, clinical decision support software falls outside FDA device regulation if it displays patient data for clinician review, allows independent review of the basis for the recommendation and does not acquire or process medical images directly. Your governance position must thread this needle.
III.
Why file before Healthcare launch
Locks in regulatory classification. Creates a paper trail you can show health-system buyers (FDA has reviewed our position). Surfaces features that push toward device territory while you can still redesign. Costs zero in FDA fees.
IV.
The five questions you ask FDA
Does the platform meet all four Section 3060 criteria? Does generating documentation that assists 510(k) submission keep us outside device classification? Auditing of cleared devices: any device-accessory risk? Bias-flagging features: interpretation of medical data? Adaptive AI under PCCP: where is the governance/PCCP boundary?
V.
The strategic framing
Position the platform as infrastructure for the clinician's review, not as software that interprets patient data. The platform does not acquire or process patient data. The platform does not generate clinical recommendations. The platform supports the independent clinical review required under Section 3060.
VI.
Q-Sub package contents
Cover letter (type of submission, meeting type). Device description (architecture, data flows). Intended use statement. Technology overview. Regulatory position (argument for non-device classification). Numbered list of specific questions for FDA.
VII.
Counsel selection
You need an FDA regulatory specialist, not a generic healthcare attorney. Short list of firms with AI/ML device Q-Sub history. Budget guidance: $30K to $50K for Q-Sub prep, $20K to $30K for follow-up meeting if requested.
VIII.
Risks and mitigations
FDA classifies as device accessory: Pre-Sub questions surface this early; restructure before launch. FDA punts: push for informal teleconference, cite specific guidance. Adverse feedback leaks: Q-Sub responses are confidential under 21 CFR 814.9. 70-day clock blowing up launch: file six months ahead.
IX.
Timeline
T-6: retain FDA counsel. T-5: draft Q-Sub package. T-4: file with FDA. T-2 to T-1: written response. T-1: optional teleconference. T-0: Healthcare SKU launch. Compatible with Year 2 healthcare entry per the platform roadmap.
App.
Appendices
Healthcare regulatory hooks (FDA 510(k), SaMD, PCCP, HIPAA, HITECH, ONC algorithm transparency, Joint Commission). Vendor requirements (HIPAA BAA, HITRUST i1/r2, state medical privacy). Marketing claims compliance (FTC unfair-trade-practice exposure).
False or misleading claims about regulatory status are FTC unfair-trade-practice violations. You cannot say “FDA approved,” “FDA cleared” or “regulatory approved auditor” for a governance platform. You can say “aligned with SR 11-7 requirements,” “generates 510(k)-submission-ready documentation” and “HIPAA-compliant architecture, BAA available.” Bake this into the brand from day one. One wrong sales-deck slide creates years of FTC exposure.