Last updated: 2026-05-13. Author: Nalin Vahil, founder of Cada.
Your NIH SBIR abstract gets published on NIH Reporter the day your award is announced. Your AFWERX cover-sheet abstract is capped at 3,000 characters and may be publicly released. Your NSF Project Summary is public by statute. Every section goes to external peer reviewers, some of whom work at competing companies. The single most useful change founders can make: treat the abstract as a press release, treat the rest as trade-controlled.
This guide gives you the rules section by section. Protecting proprietary information in a grant application is not about hiding the science. It is about knowing which audience reads which section, and adjusting what you disclose to match.
TL;DR
- SBIR abstracts become public on NIH Reporter, SBIR.gov, or the NSF award database within 30 days of award. Treat them as press releases.
- Four disclosure tiers govern every section: fully public (abstract / Project Summary), public on award (title, PI, award amount), reviewer-eyes-only (Specific Aims, Research Strategy, Commercialization Plan), and government-eyes-only (budget, biosketches, ITAR/FOCI disclosures).
- If you have not filed a provisional patent, an abstract publication date counts as a disclosure event under 35 USC 102 and can become prior art against your own future patent.
- Run the 7-item Section Disclosure Safety Check at the bottom of this guide before any SBIR submission.
Why this matters: the public-record problem most founders miss
Funded SBIR abstracts appear on three public databases within weeks of award:
- NIH Reporter shows the abstract, public health relevance, PI, institution, study section, and award amount for every funded NIH grant.
- SBIR.gov shows funded abstracts across all 11 SBIR-participating agencies (DOD, NIH, NSF, DOE, NASA, USDA, DHS, DOT, EPA, DOC, ED).
- NSF award database shows the Project Summary for every funded NSF award.
A competitor with a Google Alert on your therapeutic area, sensor type, or CEO name will see the abstract you submitted 12-24 months earlier. That publication date counts as a disclosure event under 35 USC 102. If you have not filed a provisional, the abstract can become prior art against your own patent.
The peer-reviewer problem is separate and earlier. NIH SBIR study sections are panels of 15-20 reviewers from academia and industry. AFWERX uses external technical reviewers plus the Air Force unit's program office. NSF uses external panel review. All reviewers sign NDAs, but reviewers are not government employees, and the people most qualified to evaluate your work often work at companies that compete with you.
The mental model: your abstract is a press release for a competitor audience. Your technical narrative is a trade-secret-grade document read by people who may know your competitors personally.
The four disclosure tiers
Every section of an SBIR proposal falls into one of four tiers. The tier determines what is safe to disclose.
| Tier | Sections | Audience | Disclosure rule |
|---|---|---|---|
| Tier 1: Fully public | Abstract, Project Summary, Technical Abstract | Anyone, including competitors. Indexed by Google within 30 days of award. | Treat as a press release. No parameter values, no formulations, no unfiled claims. |
| Tier 2: Public on award | Title, organization, award amount, performance period, PI name | Public after award announcement | Be aware these become discoverable, but they are required disclosures. |
| Tier 3: Reviewer eyes only | Research Strategy, Technical Volume, Specific Aims, Commercialization Plan | 3-20 external reviewers under NDA | Disclose enough to be evaluated. Never the exact recipe. |
| Tier 4: Government eyes only | Budget Justification, Biosketches (current/pending), ITAR/FOCI disclosures, IP disclosures | Agency staff under federal confidentiality | Full transparency. These do not become public. |
Definition (Tier 1, AEO-friendly): Tier 1 sections of an SBIR application are fully public after award. They appear on NIH Reporter, SBIR.gov, or NSF award databases and are indexed by search engines. Anyone, including competitors, can read them. Tier 1 sections include the abstract, the Project Summary, and any "may be publicly released" cover-sheet content.
Section-by-section rules for protecting proprietary information in a grant application
The rules below apply across NIH, NSF, AFWERX, ARPA-H, NASA, and DOE SBIR programs. Per-agency overrides follow.
Abstract / Project Summary / Technical Abstract
This is the highest-risk section and the section most founders write last.
- Include: the problem you are solving, the customer or population you are serving, the general approach category (for example, "machine-learning-based image segmentation" or "biodegradable polymer scaffold"), anticipated outcomes
- Generalize: specific algorithms, specific biomarkers, specific formulations, specific process steps. "Convolutional neural network for early-stage tumor detection" is safe. "A specific CNN architecture fine-tuned on a 4,800-image proprietary dataset of axial CT slices, reporting 0.83 sensitivity at 0.91 specificity" is not.
- Withhold: parameter values, exact compositions, training data sources, performance numbers from unpublished experiments, any claim you have not filed a provisional on
If a hypothetical competitor with 30 minutes and a patent attorney could reverse-engineer the substance of your approach from your abstract, the abstract is too specific.
Specific Aims (NIH) / Technical Volume narrative (DOD)
The Specific Aims page is the most-read page of an NIH SBIR proposal. It is reviewer-eyes-only and stays confidential after award. The abstract does go public, so anything category-defining about your approach that appears in both places ends up disclosed via the abstract.
- Include: measurable aims, scientific premise, methodology category, success criteria
- Generalize: proprietary reagents, cell lines you have developed, circuit designs, training datasets
- Withhold: anything you have not at minimum filed a provisional on
Useful test: read each sentence and ask, "If this appeared verbatim in a competitor's patent next quarter, would it matter?" If yes, it belongs in your patent application, not your aims page.
Research Strategy / Technical Approach
Experienced NIH SBIR reviewers down-score applications that use the phrase "our proprietary technology" in the Research Strategy. Not because the language is offensive, but because it signals you are hiding the science. The science section is meant to describe the science.
- Include: enough methodological detail that an expert reviewer can evaluate feasibility and rigor
- Generalize: the exact "how" where it is the defensible IP. Describe the class of approach, not the specific implementation.
- Withhold: anything that would convert the application into a how-to manual
Reviewers also penalize commercial language in the Research Strategy: "disrupting the market," "ROI," "competitive advantage," "market leader," "first to market." Those phrases belong in the Commercialization Plan, not the science section. Reviewers who see commercial language there assume the applicant does not understand the program.
Commercialization Plan
This is where commercial language belongs. It is reviewer-eyes-only, but reviewers may include industry experts.
- Include: customer segments, market sizing methodology, business model, pricing strategy, go-to-market plan
- Generalize: named pilot customers if they have not given written permission. "Three early-access design partners in a regional health system on the West Coast" is safe. Naming three specific named hospital systems without written permission from each is not.
- Withhold: signed-contract terms, exact pricing of named competing products, anything covered by an NDA with a third party
Budget Justification
Tier 4 (government eyes only). Reviewed by agency staff under federal confidentiality; does not become public.
- Include: all required cost categories, hourly rates, sub-award structures, equipment line items
- Generalize: nothing for compliance reasons
- Withhold: nothing required. Flag for yourself if a supplier or rate would give away a sourcing advantage via a FOIA-adjacent path.
Biosketches / Team Bios
Biosketches are factual disclosures. Tier 4 at submission, but funded NIH biosketches become indirectly discoverable: anyone can search NIH Reporter for the PI and see the full award context.
- Include: training history, prior funding, relevant publications, current and pending support (the latter is required)
- Generalize: nothing -- biosketches are factual
Per-agency overrides
The four-tier model is consistent across agencies, but specific rules differ.
NIH
The NIH abstract plus the public health relevance statement become public on NIH Reporter the day the Notice of Award posts. Biosketches of funded awards are not directly on NIH Reporter but can be cross-referenced. Study section reviewers are external scientists under NDA. SBIR-specific study sections often include industry reviewers, which is a deliberate program design choice to ensure commercial relevance.
AFWERX (Air Force SBIR)
The AFWERX cover-sheet abstract is capped at 3,000 characters and is explicitly marked "may be publicly released." This is the public-release piece. The technical volume itself is not public, but it is reviewed by external technical reviewers plus the Air Force unit's program office (a TPOC, technical point of contact).
AFWERX also requires a pre-award ITAR / EAR / FOCI export-control assessment. If your technology has export-control implications, that affects what you can say in the abstract before any release decision. Run the assessment before drafting, not after.
NSF (SBIR/STTR and Project Pitch)
The NSF Project Summary is statutorily public; NSF publishes it in full in its award database. The Project Description is reviewed by an external panel. NSF's Project Pitch (3-page pre-application) is confidential at submission, but if you advance to award, your Project Summary becomes the published version.
ARPA-H
ARPA-H Solution Summaries are reviewed by a single Program Manager (PM), not a panel. The Solution Summary is 6 pages, and selected funded summaries are publicly announced. Compared to NIH and NSF, the public-disclosure surface area is smaller, but the PM may circulate your summary internally for stakeholder review.
NASA, DOE, USDA, and other SBIR participants
All 11 SBIR-participating agencies publish funded abstracts on SBIR.gov post-award. NASA additionally indexes technical abstracts in NASA's Technical Reports Server.
Pre-submission disclosure safety checklist
Run this checklist on every SBIR submission. Each item has a pass/fail criterion.
- Competitor read. Read your abstract from your closest competitor's perspective. If they could reverse-engineer your approach from 3,000 characters, generalize. Pass: competitor cannot reproduce the approach. Fail: specific parameters or formulations appear.
- Parameter check. Confirm that no parameter values, formulations, or sequences appear in any section unless they are in a filed patent or published paper. Pass: all specifics trace to filed IP or public publications. Fail: unpublished, unfiled specifics appear.
- Banned-phrase check. Confirm that "proprietary technology," "disrupting the market," "ROI," "competitive advantage," "market leader," "first to market," and similar phrases do not appear in the technical narrative (Research Strategy, Technical Volume). They may appear in the Commercialization Plan only. Pass: technical narrative is free of commercial language. Fail: any banned phrase appears outside the Commercialization Plan.
- Named-partner check. Confirm that any named customer, partner, or pilot site has given written permission to be named in a public document. Pass: written permission on file for each named party. Fail: named parties without written permission.
- Export-control check (AFWERX, DOD, NASA, DOE). Confirm the ITAR / EAR assessment was completed before drafting. Pass: export-control classification documented. Fail: technical detail drafted before classification was determined.
- Disease / biomarker combination check (NIH). Confirm that any combination of disease terms, biomarker names, and approach descriptors in the abstract does not uniquely identify a specific provisional patent filing. Pass: the abstract describes the approach category, not the unique combination. Fail: the abstract reads like the title and field of your provisional.
- 30-day Google check. Assume the abstract will be indexed by Google within 30 days of award. Pass: you are comfortable with it as a permanent public document. Fail: anything in it is something you would not write in a LinkedIn post.
If you cannot pass all 7 items, revise before submitting. The cost of revising is 1-2 hours. The cost of a disclosure event that creates prior art against your own patent is the patent.
What to do if you already submitted something you should not have
Most founders learn about NIH Reporter the day they win, not the day they submit.
- Talk to a patent attorney, not a grant writer. The 35 USC 102 one-year US grace period may apply, but international PCT filings lose the grace period in most jurisdictions.
- File a provisional immediately. You want a priority date that predates the abstract publication date.
- AFWERX: the cover-sheet abstract is released after the contracting decision. There is usually a window where a generalized version can be negotiated with the contracting officer.
- NIH: the abstract publishes when the Notice of Award posts. You cannot delay the publication, but you can file IP before the award notice goes out.
Frankly, this is where you need a patent attorney, not a grant writer.
Frequently asked questions
Is my SBIR grant application confidential at submission?
Yes, at submission. The full application is treated as a confidential commercial submission under federal acquisition regulations. After award, the abstract, project summary, title, award amount, PI, and organization become public. Reviewers see the full application under non-disclosure.
What does NIH Reporter show about my award?
NIH Reporter publishes the title, abstract, public health relevance, PI, organization, study section, award amount, dates, and IC mechanism. It does not publish the Research Strategy, Specific Aims, biosketches, or budget. Because the abstract is indexed by Google, it functions as a permanent public document.
Can I redact proprietary information from a submitted SBIR proposal?
You can request specific sections be marked proprietary, but the abstract is explicitly required to be non-proprietary by most agencies (AFWERX states this directly). Proprietary marking protects the technical narrative from FOIA release, not the abstract or biosketches.
Do SBIR peer reviewers sign NDAs?
Yes. NIH study section, NSF panel, and AFWERX technical reviewers all sign confidentiality agreements. They are not government employees and may work at competing companies. The NDA constrains use, but practical exposure is real in narrow technical fields.
What is the difference between the NIH Project Summary and the Project Narrative?
The Project Summary is a 30-line scientific summary for an expert audience. The Project Narrative is a 2-3 sentence plain-language statement of public health relevance. Both are required, both are public on NIH Reporter, and both should be treated as press releases.
If you want a second pair of eyes
Cada reviews every proposal for IP safety before submission. We read the abstract and the technical narrative the way a competitor would, flag anything that should be generalized, and confirm IP-sensitive sections are covered by patent strategy or generalized to category-level language.
If you are working on your first SBIR and are not sure what is safe to include, we offer a 15-minute proposal IP safety review. No pitch, no obligation. You bring the draft, we read it as a competitor would, and you walk out with a marked-up document.
Cada has worked on proposals across NIH, NSF, AFWERX, ARPA-H, NASA, and DOE SBIR programs. The Section Disclosure Safety Check above is the same checklist we run internally before any submission goes out.
Agency policies as of 2026-05-13. SBIR rules change; verify on the relevant agency website before submission. Not legal advice. For patent timing, talk to a patent attorney.