CRADAs: Getting Inside the Problem Before There's a Contract

Most companies trying to break into defense work think about the problem in one direction: find a solicitation, write a proposal, win a contract. That path assumes you already know what the government needs. Often you don't, and neither does the solicitation. The requirement was written by someone two or three steps removed from the soldier or operator who actually has the problem.

A Cooperative Research and Development Agreement, or CRADA, works from the other direction. It puts your engineers alongside government scientists, test facilities, data, and end users before there is money on the table, so that what you eventually build is the right thing. It is one of many avenues for joint innovation between the U.S. Government and industry, and one of the most misunderstood. This article explains what a CRADA is, what each side risks, why the arrangement can pay off for both, and how to decide whether it's the right path for your company.

What a CRADA is, and what it isn't

A CRADA is an agreement between one or more federal laboratories and one or more non-federal parties to carry out specified research or development that is consistent with the lab's mission. The authority comes from the Federal Technology Transfer Act of 1986, codified at 15 U.S.C. 3710a. The statutory definition in 3710a(d)(1) is precise about who contributes what:

• The government may provide personnel, services, facilities, equipment, intellectual property, or other resources, "but not funds to non-Federal parties."
• The company may provide funds, personnel, services, facilities, equipment, intellectual property, or other resources.

That parenthetical is the single most important fact about a CRADA. Money can flow from the company to the lab, never from the lab to the company. A CRADA is not a procurement contract or cooperative agreement; the statute says so expressly, citing 31 U.S.C. 6303 through 6305, and it isn't a grant either. It isn't governed by the Federal Acquisition Regulation. It's a binding agreement to jointly produce something, whether that is hardware, software, a test result, a data set, or new knowledge, with each side bringing what it has.

DoD implements the authority through DoD Instruction 5535.08, DoD Domestic Technology Transfer Program (September 2022), and the Army through AR 70-57, Army Technology Transfer (March 2022). Army Directive 2026-20, issued in August 2026, moved legal review of Army CRADAs to the Army Office of General Counsel, so expect that office in the approval chain for new Army agreements. DoD's instruction is also explicit that companies already holding federal contracts, Other Transaction agreements, or SBIR and STTR awards remain eligible to enter CRADAs.

CRADAs are not rare. NIST's most recent governmentwide technology transfer report counted 6,962 active DoD CRADAs in fiscal year 2020, more than half of all active federal CRADAs. The Army reports that its 21 labs and centers typically execute more than 500 new CRADAs a year, and a single organization like the DEVCOM Soldier Center carries more than 130 active agreements.

What industry gets

If the government doesn't pay you, why sign one? Because the things the government can contribute are often worth more to a young defense company than a small contract would be.

Access to the people with the problem. The hardest thing for a company outside the defense world to acquire is an accurate picture of how soldiers and operators actually work, what breaks, and what they'd trade for what. Army labs describe this benefit directly: the DEVCOM Ground Vehicle Systems Center tells prospective partners that CRADAs "offer real world operation feedback and user perspective to help tailor partner's technology." When Shift5 signed a CRADA with GVSC in 2020 to work on cyber defense for Army vehicles, the announcement emphasized access to "facilities and end users" and demonstrating cyber protections on Army vehicles. USSOCOM, in its own lessons learned on CRADAs, lists user involvement as a key to success.

Requirements before they become solicitations. A CRADA lets you learn the operational problem while it is still being defined, rather than reverse-engineering it from a statement of work. The DEVCOM C5ISR Center's satellite communications CRADAs gave partners military requirements and "real threat data," and technologies developed under them were tested at the Army's Network Modernization Experiment (NetModX) in 2020 as the Army worked to make its network more resilient for Capability Set 23.

Facilities, data, and test conditions you can't buy. LexaGene's CRADA with the DEVCOM Chemical Biological Center tested its pathogen detection system against anthrax and plague using reagents from the Army's Defense Biological Product Assurance Office. Clarifai's 2021 CRADA with the C5ISR Center's Night Vision and Electronic Sensors Directorate put its data-labeling software to work on large-scale Army electro-optical and infrared data used to train AI models. Pendar Technologies worked with the Chemical Biological Center to build a DoD and DHS chemical threat library for its handheld spectrometer, which later went into an Army and DHS detection pilot.

Intellectual property protections. The statute is more generous to industry than many companies expect:

• You keep title to inventions your own employees make (3710a(b)(2)).
• You must be offered the option of an exclusive license, in a pre-negotiated field of use, to inventions government employees make under the CRADA (3710a(b)(1)).
• Trade secrets and confidential commercial information you bring may not be disclosed (3710a(c)(7)(A)).
• Information generated under the CRADA can be protected from Freedom of Information Act release for up to five years, and since 2021 for up to 30 years with agency approval (3710a(c)(7)(B)).

A preference for small business. Lab directors must give "special consideration to small business firms, and consortia involving small business firms," and preference to companies that agree to manufacture resulting products substantially in the United States (3710a(c)(4)).

Evidence. A CRADA generates test results, user feedback, and a government technical team that has seen your product work. That evidence is what program managers and acquisition officials look for when they decide what to fund next.

What industry risks

The trade for all that access is simple: you carry the financial risk. The DoD model CRADA states that each party bears its own costs and that the government will not provide funds to the collaborator. Your engineers' time, travel, equipment, and legal review are on your books. The specific risks worth pricing in:

No guarantee of follow-on work. A CRADA does not create a buyer. USSOCOM says plainly that CRADAs "will not be used to circumvent the normal procurement process" and that there is no guarantee of follow-on contracts. The National Geospatial-Intelligence Agency's CRADA handbook calls its CRADA program and its acquisitions "separate and distinct activities." A CRADA also doesn't by itself qualify you for an SBIR Phase III award, which must derive from prior SBIR or STTR work, or for OT follow-on production, which requires a completed OT prototype. It positions you for those paths; it doesn't enter you into them. The gap between a successful lab effort and a funded program is well documented; GAO has described it as a "valley of death" between DoD's development and acquisition communities.

Time. SOCOM's published benchmarks are a useful baseline: roughly 30 days for its overarching CRADA and for individual work plans under it, and 90 to 120 days for a traditional stand-alone CRADA, depending on complexity, security, and IP. First-time partners at other organizations often report longer. SOCOM's own briefing recounts a software company that walked away after three months of negotiating a work plan.

The government's license. Even when you own an invention, the government normally receives a nonexclusive, irrevocable, paid-up license to practice it for government purposes. For government-employee inventions the lab assigns or exclusively licenses to you, the government also keeps the right, in exceptional circumstances such as unmet health or safety needs or a breach of the U.S. manufacturing commitment, to require you to license others or to license them itself (3710a(b)(1)(B) and (C)).

Publication and marking. Government scientists can publish results. The DoD model agreement gives you advance review, typically 30 days, which is enough time to file a patent but not a veto. Proprietary information you share is protected only if it is marked the way the agreement requires. Anything you don't list as background technology, or don't mark, is at risk of a dispute later.

One-sided liability terms. The model agreement has the collaborator indemnify the government for losses arising from its use of results, while the government's liability is limited to the Federal Tort Claims Act and the lab disclaims warranties.

Either side can walk. Termination on 30 days' written notice cuts both ways. A reorganization, a new lab director, or a shifted priority can end an effort you have invested in.

Export control and ownership. Most defense-relevant CRADA work is subject to ITAR or EAR controls, including deemed-export rules for foreign nationals on your team. SOCOM won't sign CRADAs with foreign-owned or foreign-controlled companies at all.

Organizational conflict of interest. This is the risk companies most often miss, and it's covered in more detail below.

What the government risks

Industry should understand the government's side, because the constraints the government operates under shape what it can agree to.

A CRADA can't be a disguised contract. Under 31 U.S.C. 6303, when the principal purpose of an agreement is to acquire property or services for the government's direct benefit or use, the government must use a procurement contract. AR 70-57 repeats the point: a CRADA won't be used when a procurement contract is the appropriate instrument. GAO has held, in the context of NASA's Space Act Agreements (Exploration Partners, LLC, B-298804, 2006), that it will review a protest alleging that an agency used a non-procurement instrument where a contract was required. A government office that uses a CRADA to get free work it should have bought is inviting exactly that challenge.

Fiscal law. The Anti-Deficiency Act (31 U.S.C. 1341) means the lab's in-kind commitments are subject to available appropriations and that open-ended government indemnities are off the table. The voluntary services prohibition (31 U.S.C. 1342) is why the agreement must be signed before the work starts. Free help offered outside a signed agreement is a problem for the government even when everyone means well.

Favoritism and ethics. CRADAs don't require competition, but AR 70-57 directs commanders to "apply fairness and sound judgment in the selection of parties." Government employees must act impartially and may not endorse a product or imply government sanction (5 CFR 2635.101(b)(8) and 2635.702). That's why your press release will be coordinated with the lab's public affairs office, and why marketing that implies Army endorsement will get you in trouble.

Procurement integrity. A partner working closely with an organization that is planning an acquisition is exposed to source selection information, which 41 U.S.C. 2102 prohibits disclosing or knowingly obtaining. Government teams have to firewall acquisition planning from CRADA work.

Conflicts of interest in later competitions. A CRADA isn't a FAR contract, so FAR subpart 9.5 doesn't govern the CRADA itself. But if a partner helps write the work statement or specifications for a later competitive buy, FAR 9.505-2 can bar that company from competing, and access to nonpublic information can create an unfair competitive advantage under FAR 9.505. (The ongoing FAR overhaul keeps subpart 9.5 largely intact; the proposed rule for Part 9 was published September 18, 2026.) The rules also recognize that in development work "it is normal to select firms that have done the most advanced work in the field," so a development partner's advantage isn't automatically unfair. How the collaboration is structured is what decides the question.

Why it works when it works

Put the two sides next to each other and the logic of the arrangement is clear. The government has problems, people who understand them, and facilities no company could afford to build. Industry has engineering talent, commercial technology, and capital that moves faster than an appropriation. A CRADA lets each side contribute what it has in surplus and get what it lacks.

For the government, the payoff is leverage. It gets private R&D investment, commercial expertise, and a paid-up license to resulting inventions without spending procurement dollars, and partner funds can even support additional lab staff outside personnel ceilings (3710a(b)(3)). GAO's review of early CRADAs found that sharing resources "helped federal laboratories and private companies accomplish the CRADA's objectives" while each met its own mission.

For industry, the payoff is being right. The most expensive mistake in defense technology isn't losing a competition; it's spending two years building something no program office will buy because it solves the wrong problem. A company that has worked alongside soldiers, seen the real threat data, and tested under real conditions writes better proposals, builds better prototypes, and knows which program office owns the problem. Industry bears the financial risk, and in exchange it gets something money often can't buy: an accurate understanding of what the user actually needs.

Shift5 is a good example of a company working several paths at once. It pitched at the Army Rapid Capabilities and Critical Technologies Office's first Innovation Day in September 2019, signed its CRADA with GVSC in April 2020, and in November 2020 received a $2.6 million prototype Other Transaction award from RCCTO for a Stryker vehicle security system. RCCTO credited the Innovation Day pitch for that award, but the CRADA gave the company direct access to Army vehicle engineers and end users while the prototype work took shape. The CRADA was never meant to be the revenue; it was the part that made the revenue easier to earn and to deliver on.

CRADAs for advising and knowledge exchange

CRADAs aren't only for hardware and software. The statute lets the company contribute "personnel, services, ... intellectual property, or other resources," and expertise is a legitimate contribution. When a government organization needs to understand a fast-moving area (artificial intelligence, commercial space, data science, counter-surveillance, or anything else where the commercial world is ahead), a CRADA can be the vehicle for joint learning, provided the arrangement is genuinely jointly beneficial.

There are real precedents:

• U.S. Transportation Command and SpaceX (2020) signed a CRADA to study concepts for moving cargo through space, a pure concept and use-case study in which, as SpaceNews reported, industry participants "are not paid but volunteer time and resources to help the government study" the problem.
• NGA and Planet (2018) used a CRADA for engagement sessions in which Planet's staff and NGA analysts worked through how NGA operates and where its gaps were, giving NGA early insight into Planet's emerging analytics.
• Rockwell Collins and the DEVCOM Aviation & Missile Center signed a CRADA in 2022 that produced the Multi-Core Processors Airworthiness Playbook, a knowledge product rather than a device, released in 2025.
• NIST's AI Consortium runs on a consortium CRADA in which members contribute technical expertise, data, and models toward shared guidelines, standards, and measurement methods.

The line to stay on the right side of is the same one discussed above. If the principal purpose is for the government to receive advice for its own use, that is a service, and services are procured. SOCOM's lessons learned note that it has refused CRADA requests that were really services. An advising CRADA holds up when it is framed as research: a defined question, a method, and a joint product such as an experiment, a technology assessment, or a co-authored technical report, with both sides learning something they didn't know before. The company's return is what it learns about the mission problem and its rights in anything invented along the way.

Three further guardrails matter for advising-type CRADAs:

1. Stay on technology, not acquisition. Helping the government understand what is technically possible is joint research. Helping draft requirements, statements of work, evaluation criteria, or acquisition strategy moves into functions FAR 7.503(d) says may approach being inherently governmental, and into the FAR 9.505-2 conflict that can bar you from the resulting competition.
2. Don't become embedded staff. Your people should work under your own supervision on scoped tasks. Continuous day-to-day direction from government supervisors starts to look like personal services (FAR 37.104).
3. Watch the group dynamic. If a government office convenes several companies to deliver consensus recommendations, the Federal Advisory Committee Act can apply. Individual input, fact exchange, and workshops are excluded (41 CFR 102-3.40); a standing panel issuing group advice is not.

How CRADAs compare with other front doors

A CRADA is one tool among several, and it's often most valuable as the first step before one of the others.

A common sequence for a new entrant is a CRADA to learn the problem and prove the technology, a demonstration or experimentation event, then a funded OT, CSO, or SBIR award. For SOF-specific paths, see our overview of SOF RACER and the rapid acquisition toolkit.

Is a CRADA right for your company?

A CRADA tends to make sense when you can afford to fund your own work for six to eighteen months, when your biggest unknown is the user's real problem rather than your technology, and when a government lab or command has something you can't get elsewhere: users, data, ranges, threat information, or specialized facilities. It makes less sense if you need revenue now, if your product is already mature and the government simply needs to buy it, or if your core IP is so sensitive that any shared development creates more risk than it's worth.

If you go forward, a few habits separate the CRADAs that pay off from the ones that stall:

• Start with the lab's Office of Research and Technology Applications. Every federal lab is required to have one (15 U.S.C. 3710(b)), and it owns the template and the process.
• Scope the joint work statement tightly. Specific tasks, deliverables, and dated milestones for each party. Vague work statements are a common source of later disputes.
• Document your background IP in a dated list attached to the agreement, share only what the work requires, and mark everything the way the agreement specifies.
• Negotiate the pieces that are negotiable: the exclusive-license field of use, publication review periods, the length of data protection, and indemnity limits.
• Keep capture and CRADA teams separate, don't accept acquisition-sensitive information, and put in writing that participation doesn't affect your eligibility for future competitions.
• Plan the transition before you sign. Identify the program office or user community that would fund the next step and the vehicle it would use. AR 70-57 itself encourages early planning of data rights when follow-on acquisition is likely.

How C2 Strategies can help

C2 Strategies' founder, Cynthia M. Carrington, served as Director of Training and Support for U.S. Army Special Operations Command and has spent more than 24 years across Special Operations, the Intelligence Community, and federal acquisition. We help companies decide whether a CRADA, an OTA, or another pathway fits their goals, identify the right lab or command, review CRADA terms and work statements, and plan the transition from collaboration to a funded award.

If you're weighing a CRADA or any other path into defense work, get in touch. You can also see our full capabilities.

This article is general information, not legal advice. Agency templates, authorities, and organizational structures change; confirm current requirements with the sponsoring lab or command before relying on them.

Sources

• 15 U.S.C. 3710a, Cooperative research and development agreements
• 15 U.S.C. 3710, Utilization of Federal technology
• 31 U.S.C. 6303, Using procurement contracts
• DoD Instruction 5535.08, DoD Domestic Technology Transfer Program (Sept. 22, 2022)
• AR 70-57, Army Technology Transfer (2 Mar. 2022)
• Army Directive 2026-20, Army Legal Services Realignment (5 Aug. 2026)
• DoD model CRADA template
• USSOCOM, Cooperative Research and Development Agreements
• USSOCOM, CRADA Crash Course (SOFIC 2019)
• DEVCOM Ground Vehicle Systems Center, CRADA overview (April 2026)
• NIST, Federal Laboratory Technology Transfer Report, FY2020
• Army Technology Transfer Program, 2024 Annual Report
• DEVCOM, Army leverages innovative industry partnerships for next-gen SATCOM capabilities (Aug. 2021)
• Shift5 awarded CRADA with U.S. Army (April 2020)
• Shift5 awarded $2.6M Army RCCTO agreement (Nov. 2020)
• LexaGene successfully completes CRADA with U.S. Army (May 2022)
• Federal Laboratory Consortium, DEVCOM CBC and Pendar award
• Army.mil, Army experts help DHS develop standoff chemical detection capability
• Clarifai partners with U.S. Army C5ISR NVESD in CRADA for large-scale data labeling (June 2021)
• Avionics International, U.S. Army signs CRADA with Collins (Oct. 2022)
• Army.mil, Industry collaboration forges new airworthiness guide (2025)
• SpaceNews, U.S. Transportation Command to study use of SpaceX rockets (Oct. 2020)
• FedScoop, NGA CRADA with Planet (June 2018)
• Federal Register, NIST AI Consortium notice (May 29, 2026)
• GAO, Exploration Partners, LLC, B-298804 (Dec. 19, 2006)
• GAO/RCED-95-52, Technology Transfers: Benefits of Cooperative R&D Agreements (Dec. 16, 1994)
• GAO-23-105868, Directed Energy Weapons: DOD Should Focus on Transition Planning (Apr. 2023)
• FAR 9.505-2, Preparing specifications or work statements
• FAR 7.503, Policy (inherently governmental functions)
• 41 CFR 102-3.40, Groups not covered by FACA
• Federal Register, FAR Overhaul proposed rule, Parts 9, 27, and 47 (Sept. 18, 2026)