The negotiation is the instrument
A procurement contract arrives with its intellectual property terms already chosen. Someone decided them years ago, they sit in a numbered clause, and the conversation at award is about which of a small set of known answers applies. A prototype other transaction arrives with a draft. The Defense Innovation Unit tells companies this outright in its public guidance: nearly every aspect of an other transaction is negotiable, intellectual property included. That is why firms win unusual terms under these agreements, and why firms lose rights they never meant to give. A negotiable term is a term somebody has to actually negotiate, and the party holding the pen is usually not you.
This piece is about the negotiation rather than the terms; what the clauses say and how to write patent and data articles is covered separately in data rights under an other transaction. What follows is the process on top of that: who decides, in what order, what moves, and where firms give the outcome away before anyone opens a redline.
One framing belongs up front. Rights attach to what you deliver. Almost every hard IP fight in a prototype agreement is an argument about deliverable scope wearing an intellectual property costume, and changing what goes in the box is easier than changing the license printed on the outside.
Who is on the other side of the table
Four roles matter, and they want different things.
The agreements officer. The signature authority, and not a contracting officer. Under 32 CFR § 3.4, an agreements officer is "an individual with the authority to enter into, administer, or terminate OTs for prototype projects and make related determinations and findings." No Federal Acquisition Regulation warrant, no standard clause library, and a great deal of latitude. That latitude is the opportunity, and it means the discipline the regulation used to supply now comes from the document alone.
The program manager or technical sponsor. Where the money and the requirement live. The sponsor wants a working capability, wants to sustain it after the prototype ends, and wants to compete production later without being trapped. Every broad rights demand traces back to one of those three worries, and the sponsor is usually the only person who can agree that a narrower term settles one. An IP negotiation run entirely with the agreements officer stalls for that reason.
The consortium manager. On most Defense prototype work the government signs a base agreement with a consortium manager and members compete for project agreements underneath it. Advanced Technology International and NSTXL are two of the larger managers, both actively running consortia through 2026; ATI publicly describes more than twenty-five consortia with a combined ceiling of $7.2 billion and over 1,400 active projects. The manager is not the government, takes a fee, and administers terms it did not write.
The approving official. Section 4022 of title 10 layers written approvals as value rises, with break points at $100 million and again at $500 million and a notification to the congressional defense committees at the top tier. Most software prototypes sit far below those lines, which is why the negotiation is normally short, local, and settled by people you can get on one call.
The negotiation starts before the request goes out
By the time a draft agreement lands, a share of the outcome is already fixed. Two documents did it.
The first is the consortium base agreement, which commonly carries a standing intellectual property article. Sometimes a project agreement can vary that article and sometimes it cannot, and which is true is rarely printed on the request that reaches the membership. A term fixed at the base level is not a negotiation you are going to have, and learning that after pricing the work is expensive.
The second is the membership agreement you signed to get on the distribution list. Dues are modest and published, with NSTXL listing corporate membership from $250 to $10,000 a year depending on revenue. The cost of joining is not the point; the obligations accepted on the way in are.
Direct award is the other door. DIU solicits through commercial solutions openings, awards prototype agreements under 10 U.S.C. § 4022, and states it can do so in as few as 60 to 90 days. On that clock the intellectual property conversation is measured in days, which argues for arriving with the schedule already written.
Where leverage sits across the life of the negotiation
Editorial ranking of bargaining position by stage, drawn from statute, agency guidance, and how these agreements are structured. Relative emphasis, not measurement.
The shape of that ranking is the lesson. Bargaining position decays from the moment a request is published. It is highest when nobody has spent anything and you can still decline, and lowest after signature, when the government has a schedule to protect and every change costs both sides time. A firm that treats IP as a post-selection legal matter is negotiating from the weakest position available to it.
Do the inventory before you take a position
You cannot trade what you have not counted. Sort every asset the project will touch into three buckets before any position is worth stating.
Background. Anything predating the project or developed outside it: models, libraries, pipelines, internal tooling, datasets, evaluation harnesses. This becomes the schedule attached to the agreement, and anything left off it is exposed to the argument that it was developed under the agreement, which is far harder to win a year later than to prevent in week one.
Project-funded. What the government is paying to have built. Overclaiming background here is the fastest way to lose credibility on everything else in the schedule and to invite a challenge you have no forum to defend.
Third-party and open source. Every dependency with its license, including anything copyleft that would ship inside a delivered container. That is your disclosure obligation, not the government's discovery problem.
The schedule is half the work; the other half is engineering, and it is the half that gets skipped. If the background model and the project-funded integration layer share a repository, a build, and one shipped artifact, the schedule describes a separation that does not exist.
Ask what the government is buying, not what the clause says
The single most useful move in this negotiation costs nothing: convert every rights demand into a requirement statement before answering it.
A draft granting the government unlimited rights in all deliverables is not a considered position. It is a template, and behind it sits one of a short list of real needs: sustaining the system in year four without depending on your continued existence, competing production rather than being sole-sourced to you forever, integrating the output with a system another vendor owns, or handing the artifact to a laboratory for independent test. Sometimes the sponsor needs nothing at all and the language is what the last agreement said.
Once the requirement is named, narrower terms that satisfy it become available, and most cost far less than the original demand. It is also how the government's own negotiators prefer to work: a term written against a stated need survives a change of personnel, and a broad grant written against nothing does not.
The tradeables, and the narrower term that usually works
| What the draft asks for | The requirement usually behind it | A narrower term that meets it |
|---|---|---|
| Unlimited rights in everything delivered | Sustainment without the vendor, and the option to compete production later | Government purpose rights in the integration layer, restricted rights in the core model, and a sustainment license to a named activity |
| Source code delivery | Continuity if the firm is acquired, pivots, or fails | Escrow with defined release triggers and a verified reproducible build |
| Rights in the training data | Ability to retrain the model and audit how it was built | Corpus, annotation guidelines, and evaluation set on their own license line, with trained weights separate |
| Right to release to any contractor | Ability to hand the work to an integrator or an independent test organization | Release to named third parties under a use and non-disclosure agreement |
| A perpetual license with no stated end | Frequently nothing. Duration was never discussed | A stated term with a defined conversion at expiry, tied to a milestone |
| Rights in your existing commercial product | Fear of lock-in with no control over future price | Commercial license attached as an exhibit, with sustainment options priced at signature |
Two patterns recur in that right-hand column: split the grant by artifact instead of granting once across the project, and name who may receive the material instead of arguing about the adjective in front of "rights." Both give the sponsor something concrete to point at when leadership asks whether the government is protected.
Duration is the term nobody writes down
Draft agreements routinely use the phrase "government purpose rights" with no definition and no end date attached. Everyone in the room believes they know what it means, and what they are picturing is DFARS 252.227-7013, where the government holds those rights for "a 5-year period, or such other period as may be negotiated," after which it has unlimited rights in the data.
Picturing a clause is not incorporating it. The five-year period is a feature of the clause, not of the phrase. Use the words without the clause and you have granted rights with no expiry at all. Either reproduce the definition and the duration, or cite the clause by number and date so there is no argument about which version applies.
That clause repays reading for a second reason: it lets the parties write something else. The standard grants "may be modified by mutual agreement to provide such rights as the parties consider appropriate," subject to a floor of no less than limited rights. Under a contract that floor binds. Under an other transaction there is no floor.
If SBIR or STTR technology is going into the project, name its clock rather than assuming it travels. DFARS 252.227-7018 sets a protection period that "begins on the date of award of the contract under which the SBIR/STTR data are developed or generated and ends 20 years after that date," converting to perpetual government purpose rights afterward. Whether that protection attaches to an other transaction by force of the program's own rules is genuinely unsettled; writing it in expressly is the answer that does not depend on resolving it.
A rights concession is a price term
The most common structural error is treating intellectual property and price as separate negotiations run by separate people. They are one negotiation. A broader license is a different product and should carry a different number. Saying so moves the discussion from adjectives to arithmetic and lets the sponsor buy what they need with money rather than a term they must defend internally. Three structures do most of the work.
The step-up. Restricted rights at the prototype price; government purpose rights that arrive when the government funds a production award at or above a stated value. Rights expand when the investment expands, which is the bargain both sides would have struck anyway.
The priced option. A defined broader license, priced at signature, exercisable later at the government's election. The sponsor gets certainty about future cost, usually the real anxiety behind a demand for unlimited rights, and you are paid for the grant when it is exercised.
The funded re-implementation. If the government genuinely needs to own an implementation outright, it can fund one. Scope it, price it, and let the sponsor decide whether the requirement is worth what it costs. Often it is not.
Tie any step-up to a demonstrable milestone rather than a date. These agreements already run on milestone payment, so the machinery to verify a triggering event is in the document, and a rights change turning on a technical event outlasts one turning on a calendar.
Where the schedule actually comes from
The IP negotiation sequence, typical durations
Step five is the one most firms skip and usually the cheapest week in the sequence. Step six nobody controls: near a threshold, the calendar belongs to an approval queue.
The follow-on is the reason to be in the room
Section 4022(f) allows a prototype transaction to provide for the award of a follow-on production contract or transaction to the participants, and allows that follow-on to be awarded without competitive procedures when competitive procedures were used to select the parties and the participants successfully completed the prototype project. The statute also reaches an individual prototype or subproject carried out under a consortium, so the whole consortium effort does not have to finish for one participant's work to qualify.
Two drafting consequences follow, and they are worth more than most of the rest of the document. Define "successfully completed" against measurable acceptance criteria, because that phrase triggers the entire non-competitive pathway and satisfaction language leaves it to whoever holds the chair at the end. DIU makes this concrete: on successful completion it issues a memorandum recording the fact, and the follow-on rests on that artifact. Then say what rights the production instrument carries, because a production award made as a Federal Acquisition Regulation contract brings the standard clause set with it and your negotiated license article does not cross over on its own.
Whether the original request had to announce the possibility is a point on which readings differ; the safe practice is to put the language in both the request and the agreement.
Patents: the default you are used to is not there
Bayh-Dole runs on "funding agreements," and 35 U.S.C. § 201(b) defines that term as "any contract, grant, or cooperative agreement." An other transaction is deliberately none of the three, and the consequences run in both directions.
You have no automatic statutory right to elect title to an invention made under the project. The government has no automatic nonexclusive, paid-up license and no march-in authority under 35 U.S.C. § 203, the provision that lets an agency compel licensing where a contractor has not taken effective steps toward practical application.
Most agreements end up with a Bayh-Dole-shaped patent article anyway, because the sponsor wants the government license and the company wants title. The difference is that it has to be said out loud. Negotiate who elects title, what license the government receives and in what field, what happens on abandonment, and whether anything like march-in exists.
Four terms settled before you arrive
Audit access: 10 U.S.C. § 4022(c) and 32 CFR § 3.7 require a Comptroller General examination clause where a prototype agreement provides for total government payments in excess of $5,000,000, with access running three years past final payment. Procurement ethics: § 4022(h) treats the agreement as a Federal agency procurement for the rules in chapter 21 of title 41. Competition: § 4022(b) directs that competitive procedures be used to the maximum extent practicable. And export control, classification, and controlled unclassified information handling follow the technology, whatever the instrument is called.
Not only the Defense Department
Other transaction authority is not one authority. It is a family of statutes with different scopes, and a partner or prime working across agencies should not assume the Defense model carries over.
| Agency | Authority | What it means for the negotiation |
|---|---|---|
| Defense Department prototypes | 10 U.S.C. § 4022 | Four appropriate-use conditions at (d)(1), Comptroller General clause above $5,000,000 at (c), non-competitive follow-on production at (f). The deepest practice and the largest template library |
| NASA | 51 U.S.C. § 20113(e) | Authorizes other transactions "as may be necessary in the conduct of its work and on such terms as it may deem appropriate." Broad, and wholly dependent on the individual agreement |
| Homeland Security | 6 U.S.C. § 391 | Written by reference to the Defense section, plus a determination that a contract, grant, or cooperative agreement is not feasible or appropriate. It has carried a sunset date historically, so confirm it is live |
| Energy | 42 U.S.C. § 7256(g) | Research, development, and demonstration only, usable where standard instruments are not feasible, with the authority set to terminate September 30, 2030 |
| Health and Human Services | 42 U.S.C. § 247d-7e(c)(5) | Countermeasure development. A project expected to cost more than $100,000,000 requires a written determination by the Assistant Secretary for Financial Resources |
The commercial solutions opening authority sits alongside all of this and was amended on December 18, 2025 by Public Law 119-60. As amended, 10 U.S.C. § 3458 expressly contemplates follow-on production, including a follow-on transaction as defined in section 4022.
When the template will not move
Sometimes the answer is no, and the useful skill is telling a term the other side cannot change apart from a term nobody has questioned. A provision fixed in the consortium base agreement will not move for you. A sentence carried forward because nobody removed it moves the moment someone asks why it is there.
When a grant genuinely cannot be narrowed, narrow the deliverable instead. Rights attach to what is delivered, so changing the delivery changes the exposure without touching the article anyone is defending. Deliver the interface specification and host the model. Deliver the trained artifact and retain the training pipeline. Deliver the evaluation results and retain the evaluation set. Each is a normal scope decision, easier to agree in a technical conversation than a legal one.
Keep walking away on the table. It is the only leverage that does not decay, and it disappears the moment you price the work as though the terms were acceptable.
The position has to survive the project
A negotiated position nobody maintains is a paragraph, not a right. The work after signature is small and unglamorous, and it is where good outcomes are lost.
- Markings applied in the build pipeline, not by hand — legends in a design document do not carry into a container image or a model checkpoint
- Repository segregation verified before every delivery, not asserted once at award
- A third-party and open-source inventory refreshed at each delivery
- Records retained through the audit window — where the Comptroller General clause applies, access runs three years past final payment
- Every modification treated as an intellectual property event, with the schedule updated first
- Follow-on acceptance criteria tracked as a deliverable — "successfully completed" is a factual finding, and its evidence accumulates during performance
Bottom line
The freedom to write your own intellectual property terms and the absence of a safety net are the same fact. The regulation an other transaction displaces did two jobs: it set the default allocation, and it supplied the procedure for arguing about that later. Firms replace the first and forget the second, and the loss surfaces years on, when there is no forum and no presumption to stand on.
The negotiation rewards preparation far more than technique. Count what you have before taking a position. Ask what the government needs before answering what it asked for. Split the grant by artifact. Price the concession instead of arguing about the adjective. Change the deliverable when the license will not move. Nearly all of that happens before anyone opens a redline, which is another way of saying the outcome belongs to whoever did the work early.
Common objections, answered plainly
Isn't the IP article boilerplate the government won't touch?
Some of it is fixed at the consortium base level and will not move. A surprising amount of the rest is carried forward because nobody asked. Name the requirement the term is protecting and propose a narrower term that meets it. A sponsor who can explain to their own leadership what the government gets will usually take the narrower version.
We're the prime. Can't we push our terms down to the subcontractors?
You can, and the risk is pushing down terms the government never asked for. A flow-down broader than the head agreement costs you good partners, because the firms with valuable background intellectual property are the ones who read it and decline. Flow down what the agreement requires and price anything beyond that.
Doesn't asking for narrower rights make us look difficult?
Asking without explaining why does. Arriving with a schedule, a per-deliverable grant table, and a short statement of what the government gets under each reads as competence, and it is faster for the agreements officer than reconstructing a position from a comment thread.
Frequently asked questions
The agreements officer signs and holds the pen; under 32 CFR § 3.4 that role carries authority to enter into, administer, or terminate prototype other transactions. The person who can agree that a narrower term meets the need is usually the program manager or technical sponsor, because the requirement is theirs. On consortium work a manager runs the process but is not the government.
Generally yes as to ownership, with the government receiving a license on terms the agreement sets. DIU describes the normal outcome that way: the company retains what it created and the government licenses defined rights to use it. None of that is automatic. It happens because the agreement says so, which is why identifying pre-existing intellectual property before work starts matters more than any single clause.
On a direct award where the agency is moving quickly, days to a few weeks. DIU states it can award prototype agreements in as few as 60 to 90 days end to end, which leaves little room for a redline exchange started from scratch. Consortium project agreements commonly allow a few weeks. Where value crosses a statutory approval threshold, the approval queue sets the calendar.
Not on its own. Bayh-Dole operates through "funding agreements," which 35 U.S.C. § 201(b) defines as any contract, grant, or cooperative agreement. An other transaction is none of those, so election of title, the government license, and march-in under 35 U.S.C. § 203 exist only if the agreement creates them.
Treating intellectual property as a post-selection legal review. By then the deliverable list is set, the price assumes terms nobody tested, and the strongest argument available, declining the work, is gone. Second most common is a background schedule that does not match the repository, which turns a paper separation into an argument at delivery.