Negotiating a Joint Development Agreement: A Practical Guide

Document type: Guide Practice area: Intellectual Property — Technology Transactions Jurisdiction: United States, with cross-border notes Last reviewed: 5 September 2026

Stage 1 — The conversation before the term sheet

Have this before anything is signed, with the technical leads from your own side and, ideally, from both.

Ask four questions:

1. What are we likely to invent? Not precisely — nobody knows — but in categories. Improvements to our technology? Improvements to theirs? Integration methods? Application-specific adaptations? New chemistry, new algorithms, new mechanical designs?

2. For each category, whose technology does it belong to more naturally? Engineers answer this easily and accurately. An improvement to the polymer belongs with the polymer company; an application-specific coating formulation for implants belongs with the device company. This conversation usually produces the allocation structure in twenty minutes.

3. What will each of us need to practice the result? This surfaces the background IP question. A party that will own the foreground but needs the other's underlying technology to practice it needs a background licence that survives.

4. Where might our fields overlap? The boundary cases — veterinary, research use, food contact, adjacent markets — should be identified now and resolved in the definition.

Then draft the allocation into the term sheet. One paragraph, three sentences, before the business people write "IP developed jointly will be jointly owned" and counsel spends three weeks unwinding it.

Stage 2 — Diligence your own position first

Before committing to an allocation, confirm you can deliver it.

The assignment chain.

Do employment agreements use present-tense assignment — "hereby assigns" — rather than "agrees to assign"?
Are contractors and consultants under written agreements with present assignments? This is where the gaps are.
Are seconded personnel covered, including for inventions made at the other party's site?
For any academic collaborator, has the institution's IP policy been read?
Are prior assignments recorded with the patent office?

Fix the gaps before the project starts. Re-papering two contractors takes a week; discovering in a financing that a key inventor never assigned takes months and costs a valuation.

Your background IP.

What will you need to license, and are you free to license it? Check for existing exclusive licences, field restrictions, and upstream licence terms — a company that has granted an exclusive licence in a field cannot grant a second one.
Are there third-party components in what you will contribute, and do their licences permit this use?
For software, is there open source in the contribution, and do the licence terms create obligations for the joint work?
Do you have the rights to sublicense?

Your freedom to operate. A collaboration producing a product neither party can sell is an expensive project. Run a preliminary freedom-to-operate assessment on the intended field.

Stage 3 — Choose the allocation structure

Structure	When it fits	Key drafting
Sole ownership + licence	One party is the technology owner; the other is customer, funder, or channel	The licence terms: field, exclusivity, territory, term, sublicensing, survival
Field of use	Genuinely distinct markets	The field definitions, and the overlap
Subject matter	Each brings distinct technology; the project integrates them	The characterization mechanism
Co-ownership + full agreement	The parties genuinely want it	Consent, accounting, enforcement joinder, prosecution, abandonment, transfer

Default to sole ownership with a licence where the facts permit it. It produces the fewest disputes and the cleanest diligence position. Use field-of-use where the parties' markets are genuinely separate. Use co-ownership only with a complete agreement and only where the parties have a reason.

The negotiation moves that resolve deadlocks:

Trade ownership for exclusivity. A funder that wants ownership will often accept exclusivity in its field, which achieves its commercial objective and leaves the developer able to exploit elsewhere.
Trade breadth for term. An exclusive licence for five years followed by non-exclusive is easier to agree than perpetual exclusivity.
Trade ownership for a royalty. The party that does not own receives a share of the other's revenue in the fields it gave up.
Split by subject matter rather than by contribution. Arguing about who contributed more to an unmade invention is unproductive; agreeing that improvements to each party's technology belong to that party is not.

Stage 4 — Draft the field definitions

This is where field-of-use agreements fail, and it is worth disproportionate effort.

Define by multiple axes, not one:

Application: what the technology is used for
Product: what physical form it takes
Customer type: who buys it
Regulatory classification: how it is regulated, where relevant
Territory, if the split is geographic

Address the boundary cases expressly. For each field, state what is included and — more importantly — state what is excluded, naming the adjacent applications that will otherwise be argued about.

Include an overlap rule. Where an invention or a product falls in both fields, state what happens: joint prosecution with each owning its field's rights; ownership by one with a licence to the other; or a determination mechanism.

Include a characterization mechanism. A joint IP committee decides in the first instance; disagreements go to an independent expert with relevant technical qualifications, on a defined timetable, with the determination final. An expert determination resolves these in weeks; litigation takes years and the answer is no better.

Test the definition. Take three plausible future products and apply the definition to each. If any is ambiguous, the definition needs work. Do this with the engineers, not only with counsel.

Stage 5 — Build the background IP provisions

Three separate provisions, and parties routinely draft only the first two.

1. Identification. A schedule of each party's background IP relevant to the collaboration — patents, applications, know-how, software, materials, and data. Tedious, and the most valuable annex in the agreement.

Where full disclosure is resisted for confidentiality reasons, identify by category and reference number, with a mechanism for confirming whether a specific item is background.

2. Licence to use in the project. Non-exclusive, royalty-free, limited to performing the collaboration, terminating with the agreement.

3. Licence to exploit the result. The provision that determines whether the allocation is worth anything.

Each party grants the other a non-exclusive [or exclusive, in the licensee's Field], perpetual, irrevocable, worldwide, royalty-free [or royalty-bearing], sublicensable licence under its Background IP, solely to the extent necessary to make, have made, use, sell, offer for sale, and import products embodying the Foreground IP owned by the licensee, in the licensee's Field. This licence survives expiration or termination of this Agreement for any reason other than the licensee's uncured material breach.

Also address:

Improvements to background IP made during the project: who owns them, and is there a licence?
Third-party components in the background IP, and whether the licence can reach them
Know-how, which is background IP even though it is not registered

Stage 6 — Negotiate the commercial terms

Funding. Each bears its own; one funds the other; or shared on a split with a budget and an overrun mechanism. Resist the instinct that funding equals ownership — exclusivity in the funder's field usually achieves the commercial objective.

Milestones. Defined technically and objectively: the measurement, the threshold, the test protocol, and who performs the test. Address failure: cure period, renegotiation, termination, or a reduction in rights.

Resource commitments. Named personnel or defined full-time equivalents, for a defined period, with replacement and shortfall provisions. "Reasonable efforts to devote appropriate resources" commits nothing.

Exclusivity. May either party pursue the technology alone or with a third party, during and after? This is often what the business people care most about, and it should be drafted with the same care as the IP allocation.

Royalties, where applicable: base, rate, term, stacking, minimums, reporting, audit, and — the most litigated definition in technology licensing — combination products.

Publication. Where an academic or research party is involved, a review period of 30 to 60 days before submission, with a right to require a delay for patent filing, and a right to require removal of the other party's confidential information. Accept that publication will happen.

Stage 7 — Build the governance

A joint IP committee, meeting monthly during active development.

Its agenda, every meeting:

New invention disclosures, with contributing personnel named
Inventorship determination with counsel on the record
Ownership allocation and any characterization disputes
Filing decisions: whether, where, who prosecutes, who pays
Prosecution review — any amendment that could change the inventorship analysis
Third-party developments and freedom-to-operate
What to patent and what to keep as a trade secret

Item 5 is the one that requires discipline. A claim amended to incorporate the other party's contribution creates co-ownership in a patent that was solely owned. Patent counsel should flag every substantive amendment before filing.

The invention disclosure form should capture: the invention; the date of conception; every contributor and what each contributed; background IP used; and the allocation category.

Project governance, separately: a steering committee for technical direction, budget, and schedule, with escalation to senior management.

Stage 8 — Draft the survival provisions

Address each expressly:

Licences. Do they survive termination? For most collaborations they must. The usual formulation: survive except on termination for the licensee's uncured material breach.
Prosecution and maintenance of jointly relevant patents, and what happens if one party declines to fund a jurisdiction — typically an offer to the other, which may take over and own that jurisdiction.
Confidentiality, with a term long enough to be useful.
Data, including regulatory data and rights of reference.
Non-compete and exclusivity after termination.
Payments, with reporting and audit.
Materials and samples.
Change of control. What happens if one party is acquired by the other's competitor: suspension of information rights, a termination right, or a buyout. Negotiate this at the outset, when neither party knows which side of it they will be on.

Stage 9 — Counterparty-specific issues

A university. Read the IP policy. Negotiate the publication review period. Expect the institution to retain research and teaching rights and to resist broad exclusivity. Check for federal funding and the reporting and march-in obligations that follow. Confirm the institution actually owns what its researchers invent — the assignment analysis applies to them too.

A government contractor. Rights in technical data and software developed under government funding are governed by acquisition regulations and follow the data. Run the analysis before the work starts, because a commercial collaboration touching government-funded development can give the government rights in the result.

A customer. Its standard form assigns everything to it. Counter: the customer owns deliverables and anything specific to its application; the supplier owns its underlying technology and all improvements to it; the customer gets a broad licence to use.

A competitor. Scope the collaboration narrowly and document the scope. Restrict information flows outside the field, with an information protocol and clean teams. Have antitrust counsel review before signing.

A startup. The larger party should check whether its standard form makes the startup unfinanceable. The startup should expect diligence on its assignment chain, its open source position, and its freedom to operate — and should fix those before the negotiation, because they will be found.

Cross-border. Inventorship and ownership rules differ; some jurisdictions require a first domestic filing or a foreign filing licence; employee invention compensation is mandatory in several countries and cannot be contracted away; and export control classification should be run before technical data crosses a border.

Running the project: what counsel does after signing

The agreement is the beginning. Most of the value counsel adds in a joint development is in the two years after signing, and most of the failures happen there.

Month one — set up the machinery.

Convene the first joint IP committee meeting and establish the cadence
Distribute the invention disclosure form and train the engineers on when to use it
Confirm every person working on the project — employees, contractors, secondees, interns — is covered by a present-tense assignment
Establish the background IP reference process: how a party confirms whether a specific item is background
Set up the project record: a shared, dated repository of technical reports, meeting minutes, and disclosures
Brief the engineers on what not to do: no use of the other party's confidential information outside the project; no incorporation of open source without review; no publication without the review process

Ongoing — the recurring items.

Invention disclosures. Reviewed at every committee meeting, with inventorship determined and recorded contemporaneously. The value of the record is that it exists when the invention was made, not when a dispute arises.

Claim amendments. Patent counsel flags every substantive amendment for inventorship review before filing. This is the single most important ongoing discipline and the one most often neglected.

Field characterization. Each invention allocated to a field at the time of disclosure, with the reasoning recorded. Disagreements escalated promptly rather than deferred.

Open source scanning. For software deliverables, periodic scanning and review of any component introduced, with the licence obligations assessed.

Publication requests. Routed through the review process, with the clock started and the patent filing decision made within it.

Milestone certification. Documented at the time, with the test results, rather than reconstructed at payment.

Resource compliance. Whether the committed personnel are actually assigned. This is where collaborations quietly fail, and the record matters if a shortfall becomes a dispute.

At each anniversary.

Review the background IP schedules for additions
Confirm the assignment chain covers anyone who joined the project
Review the field definition against the products that have actually emerged
Confirm the licence provisions still match the commercial reality
Assess whether the collaboration is delivering, and whether either party's strategy has changed

And the item that matters most at the end. Six months before termination or expiry, review the survival provisions and confirm that each party has what it needs to continue: licences that survive, data it may use, patents it may prosecute, and materials it may keep. Discovering at termination that a needed licence does not survive is a problem with no solution.

A worked negotiation: Fennimore and Oyelaran

The parties. Fennimore Photonics, a laser component maker, and Oyelaran Diagnostics, which builds analytical instruments. They will develop a laser module optimized for Oyelaran's next-generation instrument.

The opening position. Oyelaran's standard development agreement assigns all IP to Oyelaran, on the ground that it is paying. Fennimore's counsel, Priya Venkataraman, cannot sign it: the module will incorporate improvements to Fennimore's core laser technology, and assigning them would transfer Fennimore's business.

Stage 1 — the technical conversation. Venkataraman convenes Fennimore's chief technology officer and asks the four questions. The answer is clear: the project will produce (a) improvements to Fennimore's laser cavity design, (b) improvements to Oyelaran's optical detection algorithms, and (c) a module integration design specific to Oyelaran's instrument.

The allocation writes itself:

(a) belongs to Fennimore — it is an improvement to its core technology
(b) belongs to Oyelaran — same logic
(c) belongs to Oyelaran, with a licence back to Fennimore for non-competing applications

Stage 2 — Fennimore's own diligence finds that two of the four engineers assigned to the project are contractors under agreements with "agrees to assign" language. Re-papered in five days, before the project starts.

Stage 3 — the negotiation. Oyelaran resists, wanting ownership of everything. Venkataraman offers the trade that resolves it: exclusivity rather than ownership.

Fennimore owns improvements to its laser technology; Oyelaran receives an exclusive licence to those improvements in the Diagnostic Instrument Field for [five] years, then non-exclusive, royalty-free.

Oyelaran's commercial objective — that its competitors cannot get the improved laser — is achieved. Fennimore retains its technology and can exploit it in industrial, defence, and telecommunications applications.

Stage 4 — the field definition. "Diagnostic Instrument Field" is defined by application (in vitro diagnostic and analytical instrumentation), by regulatory classification, and with express exclusions for industrial process monitoring, environmental testing, and research instrumentation not used for diagnosis. The engineers test it against three planned products.

Stage 5 — background IP. Both schedules built. And the provision Venkataraman insists on: Fennimore grants Oyelaran a licence to its background laser IP, in the Diagnostic Instrument Field, sufficient to make and sell instruments incorporating the Module, surviving termination. Without it, Oyelaran would own a module design it cannot practise.

Stage 7 — governance. A joint IP committee, monthly. In month fourteen it earns its existence: a claim amendment proposed by Oyelaran's patent counsel would incorporate a limitation contributed by a Fennimore engineer, creating co-inventorship in a patent Oyelaran was prosecuting alone. The committee catches it, the amendment is redrafted, and the patent stays solely owned.

Stage 8 — change of control. Fennimore negotiates: if Oyelaran is acquired by a laser manufacturer, Fennimore may terminate the exclusivity and the information rights are suspended pending a 90-day resolution period. Oyelaran accepts, because it wants a symmetrical provision.

Two years later Oyelaran is acquired — by a diagnostics company, not a laser maker. The provision does not trigger, and both parties are glad it exists anyway.

When a dispute arises

Joint development disputes cluster into a small number of shapes, and the response differs for each.

Characterization disputes — is this invention foreground or background, Field A or Field B, an improvement to your technology or to ours?

Response: use the mechanism. A joint IP committee determination, then an independent technical expert on a defined timetable. These are technical questions with technical answers, and a qualified expert resolves them in weeks. Litigating them takes years and produces a worse answer from a decision-maker with no technical background.

If the agreement has no mechanism, propose one. Parties in a live dispute will frequently agree to an expert determination because both prefer speed.

Inventorship disputes — should our engineer have been named?

Response: this is a factual question decided on contemporaneous evidence. Assemble the notebooks, disclosure forms, emails, and meeting records for the relevant period. Correction is available and is easier before issuance and easier without any suggestion of deceptive intent, which is why an error found and corrected promptly is very different from one asserted years later in litigation.

Scope disputes — is this product within the licence field?

Response: apply the definition, and if it does not resolve the question, that is a drafting failure to be negotiated rather than argued. A licensor asserting that a licensee's product is outside the field is threatening the licensee's business, and the commercial consequences usually drive a resolution.

Performance disputes — resources not provided, milestones missed, information withheld.

Response: notice under the agreement, promptly and in the required form. Parties forgive these informally and then find they have waived a right they now need. Send the notice, and then negotiate.

Confidentiality and use disputes — the other party's people used our information outside the project.

Response: the hardest to prove and the most damaging. Preserve the evidence, check the residuals clause, and assess whether the information was genuinely confidential and identified as such.

The structural point. Almost all of these are cheaper to resolve than to litigate, and the agreement's dispute mechanisms — a committee, an expert, an escalation ladder to senior executives — resolve most of them if the parties use them. The collaborations that end in litigation are usually the ones where nobody used the mechanism until the relationship had already failed.

The document set

A joint development is more than one agreement, and the pieces should be planned together.

The joint development agreement itself, covering: the project and its scope; funding and milestones; resource commitments; background IP identification and licensing; foreground IP allocation; prosecution and enforcement; confidentiality; governance; term, termination, and survival; representations and warranties; indemnities and limitation of liability; and dispute resolution.

Schedules and annexes, which carry more of the substance than the body:

Background IP schedules, one per party
The statement of work or development plan, with milestones, deliverables, and acceptance criteria
The budget, where costs are shared
Named personnel, where resource commitments are specific
Field definitions, if long enough to warrant a separate schedule
Approved subcontractors

Ancillary agreements:

A confidentiality agreement, if one is not already in place — and check whether an earlier NDA governs and whether it should be superseded
Material transfer agreements, where physical materials pass
Software licence terms, for any software contributed or delivered
A secondment agreement, where personnel work at the other's site, addressing direction of work, benefits, invention assignment, confidentiality, and site rules
A quality or technical agreement, in regulated industries
A supply agreement, where one party will supply components for the resulting product — negotiate this at the same time, because a party that develops jointly and then negotiates supply from a position of dependence gets poor terms
A commercialization or distribution agreement, or at least a term sheet for one

The sequencing point. The supply and commercialization terms are where the collaboration's value is ultimately realized, and negotiating them after the development is complete puts the party that needs them in a weak position. Agree at least a binding term sheet for the downstream arrangements at the same time as the development agreement, even if the definitive documents follow.

Warranties, indemnities, and liability

These provisions get less attention than the IP allocation and generate real exposure.

Warranties each party should give:

Ownership and authority: it owns or has the right to license the background IP it contributes, and entering the agreement does not breach any other obligation
No known infringement: to its knowledge, the contributed background IP does not infringe third-party rights. Note the knowledge qualifier — an unqualified non-infringement warranty on technology is rarely appropriate and is heavily negotiated
No conflicting exclusive grants in the fields being licensed
Personnel: everyone working on the project is under an obligation to assign inventions and to maintain confidentiality
Open source: for software contributions, disclosure of open source components and their licences
Compliance: with export control and applicable law

Indemnities:

Third-party IP infringement claims arising from a party's background IP should be indemnified by that party. This is the core allocation and it is usually accepted.
Claims arising from the foreground IP are harder. Neither party can sensibly indemnify the other against infringement by an invention neither has yet made. The workable approach is a shared cost mechanism, a duty to cooperate in defence, and a right for either party to seek a licence or to design around, with the costs shared.
Product liability should follow the party that manufactures and sells, with the other indemnified for claims arising from its contribution's defect.
Employment and personnel claims by each party for its own people.

Limitation of liability. A cap, usually expressed as a multiple of amounts paid or a fixed sum, with the standard carve-outs: breach of confidentiality; IP infringement indemnity; gross negligence and wilful misconduct; and, sometimes, breach of the exclusivity or field restrictions.

Consequential damages excluded mutually, with the same carve-outs.

The provision worth arguing about: whether breach of the field restrictions is inside or outside the cap. A party that exploits outside its field has taken the other's core business, and a claim capped at development costs is no remedy. Carve it out, or provide a separate remedy — injunctive relief expressly acknowledged as appropriate, and a liquidated amount.

A negotiation timetable

Week	Task	Who
−2	The four-question conversation with the technical leads	Counsel + CTO
−1	Draft the allocation paragraph for the term sheet	Counsel
−1	Own-side diligence: assignment chain, background IP freedom, open source	Counsel
0	Term sheet signed with the allocation in it	Business + counsel
1	First draft circulated	Drafting party
1–2	Background IP schedules built	Both, technical + IP counsel
2	Field definitions drafted and tested against three planned products	Counsel + engineers
2–3	Commercial terms: funding, milestones, resources, exclusivity	Business
3	Warranties, indemnities, liability	Counsel
3–4	Survival, change of control, dispute resolution	Counsel
4	Downstream term sheet: supply and commercialization	Business
4	Antitrust review, if the parties compete	Antitrust counsel
5	Ancillary documents: secondment, materials transfer, software terms	Counsel
5	Execute
6	Set up the machinery: IP committee, disclosure forms, assignment confirmation, project record, engineer briefing	Counsel

The two entries that determine the quality of the agreement are the week −2 conversation, which produces the allocation, and the week 2 field definition testing, which determines whether the allocation survives contact with actual products.

Quick reference

Before the term sheet, ask the technical leads four questions: what will we invent, whose technology does each category belong to, what will each of us need to practise the result, and where might the fields overlap. The answers produce the allocation.

Diligence your own side first. Present-tense assignments for employees, contractors, and secondees; freedom to license your background IP; open source position; and a preliminary freedom-to-operate view.

Default to sole ownership with a licence. Use field-of-use where markets are genuinely distinct. Use co-ownership only with a complete agreement fixing consent, accounting, enforcement joinder, prosecution, abandonment, and transfer.

The negotiation moves that break deadlocks: trade ownership for exclusivity; breadth for term; ownership for a royalty; and allocate by subject matter rather than by contribution.

Test the field definitions against three real products, with the engineers.

Three background IP provisions, not two: identification, a licence to use in the project, and a licence to exploit the result that survives termination.

Build the governance and use it. A joint IP committee reviewing disclosures monthly, with counsel flagging every claim amendment that could change inventorship.

Draft survival at the outset, including what happens if one party is acquired by the other's competitor.

And negotiate the downstream supply and commercialization terms at the same time, because the party that needs them after the development is finished will not get good ones.

Errors that recur

Before signing.

"IP developed jointly will be jointly owned" in the term sheet, written by business people before counsel is involved
No conversation with the technical leads about what will actually be invented
No own-side diligence, so a party commits to an allocation it cannot deliver because a contractor never assigned
Field definitions drafted by lawyers alone and never tested against real products
Background IP schedules omitted because they are tedious
A background licence covering the project but not the exploitation of the result
Downstream supply and commercialization terms deferred

During the project.

No joint IP committee, or one that meets twice and stops
Invention disclosures not recorded contemporaneously, so inventorship becomes a memory contest
Claim amendments filed without inventorship review — the failure that silently converts sole ownership into co-ownership
New personnel added without confirming their assignment coverage
Open source introduced into deliverables without review
Publications submitted without the review period
Resource shortfalls tolerated informally until they are a pattern
Characterization disagreements deferred rather than escalated

At the end.

Survival provisions read for the first time at termination
A needed licence that does not survive
Data or regulatory rights that cannot be used
Patents nobody agreed to continue prosecuting, abandoned by default
A change of control that the agreement never addressed

The through-line. Almost every one of these is a failure to do something small at a moment when it was easy, and to discover it at a moment when it is not. The agreement is worth what the administration behind it is worth, and counsel who negotiates a good document and then disappears has delivered half the work.

Advising the smaller party

Where one party is materially larger, the negotiation is asymmetric, and the smaller party's counsel needs a different playbook.

Know what you cannot give away. For a technology company, the improvements to its core technology are the business. Assigning them, or granting an exclusive licence to them across all fields, ends the company — and no amount of development funding compensates. Identify this line before the negotiation and do not cross it.

Understand the large party's actual objective. It is almost never ownership for its own sake. It is usually: certainty of supply; exclusivity against its competitors; freedom to practise without a future royalty demand; and the ability to continue if the small company fails. Each of those can be given without transferring ownership — through a supply commitment, field exclusivity, a paid-up licence, and an escrow or step-in right.

Offer the step-in right early. A large customer's real fear is that the small supplier disappears. A source code escrow, a technology escrow, or a licence that becomes broader on the supplier's insolvency addresses it directly and is far cheaper than ownership.

Protect financeability. An investor will examine whether the company still owns its core technology and whether it can sell to other customers. A term sheet that grants field exclusivity in the company's largest addressable market makes the next round harder, and counsel should model that before agreeing.

Watch the diligence burden. A large counterparty's process will consume the small company's technical leadership for weeks. Negotiate a scoped diligence exercise with a defined document list and a deadline.

Watch the payment terms. Development milestones paid ninety days after acceptance, with acceptance in the large party's discretion, can be a cash flow problem that outweighs the deal's value. Negotiate defined acceptance criteria and a deemed-acceptance period.

And use the asymmetry. A large company that has committed engineering resources and told its own management the project is happening has more invested in the outcome than the term sheet suggests. The smaller party's leverage is highest just before signing and again just before a milestone, and counsel should know which of those moments they are in.

Advising the larger party

The mirror image, and it is not simply the opposite.

Your standard form may kill the deal, or the counterparty. A development agreement assigning all IP to the customer is unsignable by a technology company whose improvements to its own core technology would be transferred. Have a second form — or a set of pre-approved deviations — for counterparties whose technology is the point of the collaboration.

Ask what you actually need, and get that instead of ownership. Usually: exclusivity in your field for a period; a paid-up, perpetual, sublicensable licence sufficient to make and sell the product without a future royalty demand; continuity if the supplier fails; and freedom from a claim that the product infringes the supplier's own patents. All four are achievable without transferring ownership, and they are easier to get than ownership is.

Diligence the counterparty properly, and early. The assignment chain, the open source position, the upstream licence terms, and whether it has granted an exclusive licence in your field to someone else. Finding an exclusivity conflict after signing is expensive.

Consider whether the counterparty will survive. A development partner that runs out of money mid-project leaves you with a partial deliverable and an ownership dispute with a receiver. Escrow, step-in rights, and milestone-based funding that does not front-load your exposure are the responses.

Do not over-specify the resource commitment in a way you will breach. Large companies commit named engineers and then reassign them, which is a breach the counterparty will raise.

And be realistic about the field definition. Your business people will want the field defined as broadly as possible; a field so broad that it captures the counterparty's entire market will either be refused or will end the relationship in year three when the counterparty realizes what it agreed to. A durable collaboration needs both parties to have a business afterwards.

A short note on trust

Every provision in this guide exists because collaborations that began in good faith ended badly, and the drafting is an attempt to make the ending survivable. It is worth saying plainly that the drafting is not a substitute for the relationship.

The agreements that work are the ones where both parties actually want the other to succeed, where the technical teams respect each other, and where the commercial logic holds for both sides over the life of the project. No allocation clause rescues a collaboration in which one party has concluded it made a bad deal, and counsel who believe otherwise are overestimating what documents do.

What the drafting achieves is narrower and still valuable: it prevents the good-faith collaboration from failing over a question nobody thought to ask; it gives the parties a mechanism to resolve an honest disagreement quickly; and it ensures that when the collaboration ends — as most do, on schedule or otherwise — each party leaves with what it needs to continue.

The practical implication for how to negotiate. Raise the difficult questions early and directly, in terms the business people understand. "If we invent something that improves your polymer and our device, who owns it?" is a question both sides can answer in a meeting. Deferring it because it is uncomfortable does not make it easier; it makes it a dispute between people who by then have positions.

And the test of a good agreement is not whether it protects your client in every scenario. It is whether both parties, reading it in year three when something has gone wrong, find that it answers the question and gives them a way forward. An agreement that only one party can live with will produce a collaboration that only one party works at.