Joint Development Agreements and Co-Owned Intellectual Property: Who Owns What When Two Companies Build Together

Document type: Article Practice area: Intellectual Property — Technology Transactions Jurisdiction: United States Last reviewed: 5 September 2026

The default rule that nobody wants

Two companies collaborate. Engineers from both contribute. A patentable invention results, and both companies' employees are named inventors. Each assigns to its employer. The patent is now co-owned.

Here is what co-ownership means under 35 U.S.C. § 262:

In the absence of any agreement to the contrary, each of the joint owners of a patent may make, use, offer to sell, or sell the patented invention within the United States, or import the patented invention into the United States, without the consent of and without accounting to the other owners.

Read that again. Each co-owner may:

Practice the invention freely
License it to anyone, including the other co-owner's direct competitor
License non-exclusively at any royalty, or for nothing at all
Do all of this without the other's consent
And keep every dollar, with no obligation to account

One co-owner can destroy the patent's value unilaterally, by licensing it broadly and cheaply, and the other has no remedy.

And neither can enforce it alone

The second half of the problem is procedural, and it is worse.

All co-owners must join in an infringement suit. A co-owner that refuses to join cannot generally be forced to, and the action cannot proceed without it.

Ethicon, Inc. v. United States Surgical Corp., 135 F.3d 1456 (Fed. Cir. 1998) is the case every technology lawyer should know. An accused infringer, facing a patent suit, located an omitted co-inventor of one claim, obtained a retroactive license from him, and had the suit dismissed. The co-inventor's license was a complete defense, because a co-owner may license without the other's consent, and the plaintiff could not proceed without a co-owner who had licensed the defendant.

Schering Corp. v. Roussel-UCLAF SA, 104 F.3d 341 (Fed. Cir. 1997) held that a co-owner cannot be involuntarily joined as a plaintiff under the rule permitting involuntary joinder, confirming that a reluctant co-owner is a veto.

STC.UNM v. Intel Corp., 754 F.3d 940 (Fed. Cir. 2014) confirmed the rule in a case where the co-owner declined to join and the patent owner sought to compel it. The suit could not proceed.

The practical consequence for a joint development agreement: if the parties leave the patent co-owned without an agreement, they have created an asset that neither can enforce and either can devalue. Co-ownership is not a compromise. It is a failure to decide.

Joint inventorship: how it happens

Co-ownership arises from joint inventorship, and joint inventorship arises more easily than engineers expect.

35 U.S.C. § 116 provides that inventors may apply jointly even though they did not physically work together or at the same time, did not each make the same type or amount of contribution, and did not each make a contribution to the subject matter of every claim.

The elements of joint inventorship:

Contribution to the conception of at least one claim — conception being the formation of a definite and permanent idea of the complete and operative invention
Some collaboration or connection between the inventors
A contribution that is not insignificant in quality when measured against the full invention
More than merely explaining well-known concepts or the state of the art

What does not make someone an inventor:

Building or testing what another conceived, however skilfully
Suggesting a desired result or a problem to be solved
Providing funding, facilities, or materials
Supervising
Explaining existing knowledge

The trap. A single engineer's contribution to a single claim makes that person a co-inventor of the entire patent, and their employer a co-owner of the whole thing. Contribution to one claim confers ownership of all of them.

And inventorship is a question of fact determined by the claims as ultimately allowed, which means it can change during prosecution. A claim amended to incorporate a feature contributed by the other party's engineer creates co-inventorship where none existed at filing.

The related problem: whether the company owns the invention at all

Co-ownership presupposes that each company owns its employees' inventions, and that is not automatic.

An invention belongs to the inventor unless assigned. Board of Trustees of the Leland Stanford Junior University v. Roche Molecular Systems, Inc., 563 U.S. 776 (2011) applied this rule in a federally funded research context, holding that federal funding legislation does not automatically vest title in the contractor — the invention belongs to the inventor unless and until assigned, and a researcher who had signed a present assignment to a company had already transferred his rights before his university obtained a later promise to assign.

The drafting lesson from that case is now standard practice: an agreement in which an employee "agrees to assign" is a promise to make a future assignment, which may lose to an intervening present assignment. An agreement stating that the employee "hereby assigns" effects a present transfer of future inventions.

FilmTec Corp. v. Allied-Signal Inc., 939 F.2d 1568 (Fed. Cir. 1991) established the principle, and Omni MedSci, Inc. v. Apple Inc., 7 F.4th 1148 (Fed. Cir. 2021) illustrates its limits, holding that a university policy providing that inventions "shall be the property of the university" was a statement of intended disposition rather than a present automatic assignment — so the inventor retained title and could assign elsewhere.

The practical instruction. Before entering a joint development agreement, confirm that each party actually owns what its people invent: present-tense assignment language in employment agreements; contractors under written agreements with present assignments; and, for academic collaborators, the institution's policy examined rather than assumed.

Structures that work better than co-ownership

The central drafting decision in a joint development agreement is how to allocate foreground intellectual property, and there are four approaches.

1. Sole ownership by one party, with a license to the other

The cleanest structure. One party owns everything developed under the agreement; the other receives a license — exclusive or non-exclusive, in a defined field, for a defined term, with defined sublicensing rights.

When it fits: where one party is clearly the technology owner and the other is a customer, a funder, or a channel; where one party will commercialize and the other will not; and where the parties' fields of use are cleanly separable.

The negotiation is then about the license, which is a much better negotiation than one about ownership: field, exclusivity, territory, term, sublicensing, royalties, improvements, and what happens on termination.

2. Field-of-use allocation

Each party owns the foreground IP in its own field, with a license from the other in the fields it does not own.

When it fits: where the parties operate in genuinely distinct markets — a materials company and a medical device company developing a coating, each owning the result in its own application space.

The drafting requirement is a precise field definition, and this is where these agreements fail. "Industrial applications" and "medical applications" seem clear until a product is both. Define the fields by application, by product, by customer type, and by regulatory classification, and address the overlap expressly.

3. Subject-matter allocation

Ownership follows what the invention relates to: improvements to Party A's contributed technology belong to A; improvements to B's belong to B; and genuinely new inventions belong to whoever conceived them, or to one party by agreement.

When it fits: where each party brings a distinct technology and the collaboration integrates them.

The difficulty is characterization — deciding whether an invention is an improvement to A's technology or a new invention — and the agreement needs a mechanism: a joint IP committee, an expert determination, or a default rule.

4. Co-ownership with an agreement that fixes the defaults

Sometimes the parties genuinely want co-ownership. Section 262 permits agreement to the contrary, and the agreement must address every default:

Consent to license. Each co-owner agrees not to license without the other's consent, or on defined terms
Accounting. Each agrees to share revenue from licensing, on a stated split
Enforcement. Each agrees to join any suit the other brings, at the other's cost, and not to unreasonably withhold consent — this is the provision without which the patent is unenforceable
Prosecution. Who files, in which countries, who pays, and who controls; and what happens if one party declines to fund a country
Abandonment. If one party wishes to abandon, it must offer its interest to the other first
Transfer. Restrictions on assigning a co-ownership interest, with a right of first refusal
Practice. Confirmation that each may practice without accounting, if that is intended
Marking, maintenance fees, and administration

Co-ownership with a complete agreement is workable. Co-ownership without one is the failure described at the start of this article.

Background IP: the other half of the problem

Foreground IP gets the attention; background IP causes as many disputes.

Background IP is what each party brings — existing patents, know-how, software, materials, and data. The foreground development will almost certainly use it, and the resulting product may be unusable without it.

The essential provisions:

Identification. A schedule listing each party's background IP relevant to the collaboration. This is tedious and it is the single most valuable annex in the agreement, because a dispute about whether something is background or foreground is a dispute about ownership.

Retained ownership. Each party retains its background IP. Uncontroversial and always stated.

License to use it in the project. Each grants the other a license to use its background IP for the purpose of performing the collaboration. Limited, non-exclusive, royalty-free, and terminating with the agreement.

License to exploit the result. This is the provision that matters and is most often missed. A product embodying the foreground IP may be unusable without a license to the background IP. A party that owns the foreground and has no background license owns something it cannot practice.

The provision should state: which background IP is licensed for exploitation; the field, territory, and exclusivity; whether royalties apply; whether it is sublicensable; and whether it survives termination.

Improvements to background IP. If Party A's engineers improve Party B's background technology during the project, who owns the improvement? State it. The common answers are that improvements to a party's background IP belong to that party, with a license to the other; or that they are foreground IP allocated under the general rule.

A caution about the identification schedule. Parties resist producing it because it requires disclosing what they have. The resistance should be overcome, because the alternative — arguing three years later about whether a technology was background or foreground — is far worse. Where confidentiality is the concern, identify by category and reference number rather than by full technical disclosure.

Copyright and software

Where the collaboration produces software, documentation, or other copyrightable material, a different set of rules applies.

Copyright vests in the author, and for a work made for hire, in the employer. 17 U.S.C. § 201 provides that the authors of a joint work are co-owners of the copyright.

A joint work is one prepared by two or more authors with the intention that their contributions be merged into inseparable or interdependent parts of a unitary whole.

Joint authorship is harder to establish than joint inventorship. Childress v. Taylor, 945 F.2d 500 (2d Cir. 1991) requires that each putative author contribute independently copyrightable expression and that both intend to be joint authors. Aalmuhammed v. Lee, 202 F.3d 1227 (9th Cir. 2000) added that a joint author must exercise control over the work — a superintendence factor that keeps contributors from becoming co-owners.

The co-ownership rules for copyright differ from patent, in one important respect:

Each co-owner may use and license non-exclusively without the other's consent — like patent
But each co-owner must account to the other for profits. Unlike patent, there is a duty to account
An exclusive license requires all co-owners
Either co-owner may sue for infringement without joining the other — unlike patent

So joint copyright ownership is less catastrophic than joint patent ownership, but it is still a poor outcome, and the same allocation approaches apply.

For software specifically, address:

Ownership of the source code, module by module where the parties contribute separately
Open source components: what licences apply, whether copyleft obligations attach, and who bears the compliance obligation
Third-party components and their licence terms
Escrow of source code where one party depends on the other's continued performance
Derivative works and who may create them
Documentation, which is separately copyrightable and frequently forgotten

Trade secrets and data

Much of what a collaboration produces is never patented and never registered. It is know-how, process knowledge, test data, and negative results — and it is the most commercially valuable output of many collaborations.

Trade secret protection depends on secrecy, which a collaboration inherently strains.

The provisions required:

Confidentiality, with a term long enough to be useful — a five-year confidentiality term on a trade secret is a five-year trade secret
Marking and identification procedures, and a fallback for orally disclosed information
Permitted use, limited to the collaboration
Restrictions on personnel with access, and on their subsequent assignments
Return or destruction on termination, with a carve-out for archival copies and for information in backup systems
Ownership of jointly developed know-how, allocated on the same principles as patents
Ownership of data — test data, performance data, clinical data, operational data — which is frequently the most valuable output and is almost never addressed
Residuals clauses, which permit a party's personnel to use general knowledge retained in unaided memory. These are heavily negotiated and dangerous: a broad residuals clause can permit the other party to use everything its engineers remember, which for a small technology company is everything it has.

On data specifically. A collaboration generating test or performance data should state who owns it, who may use it, for what purposes, whether it may be published, whether it may be used to support a regulatory filing, and what happens to it on termination. Regulatory data in particular has independent value and its ownership determines whether a party can file without the other's cooperation.

Managing inventorship during the project

Inventorship is not a question answered once at signing. It is determined by the claims as ultimately allowed, which means it can change through prosecution, and a collaboration that does not manage it actively will discover its ownership position at the wrong moment.

The mechanism that works: a joint IP committee.

Composition. One or two representatives from each party, with technical and legal participation, and patent counsel attending.

Cadence. Monthly during active development, quarterly thereafter.

Standing agenda.

New invention disclosures since the last meeting, with the contributing personnel identified by name and date
Inventorship determination for each, with counsel's analysis on the record
Ownership allocation under the agreement's rules, and any characterization disputes
Filing decisions: whether to file, where, and who prosecutes and pays
Prosecution review: pending amendments, and whether any amendment changes the inventorship analysis
Third-party developments: competitor filings, freedom-to-operate issues
Trade secret decisions: what will be patented and what will be kept confidential

Item 5 is the one that requires discipline and is most often skipped. A claim amended during prosecution to incorporate a limitation contributed by the other party's engineer creates co-inventorship, and therefore co-ownership, in a patent that was previously solely owned. Patent counsel should flag every substantive amendment to the committee before it is filed.

The invention disclosure form should capture: the invention; the date of conception; every person who contributed to conception, and what each contributed; whether the contribution was to a specific concept or general; what background IP was used; and whether the invention is Medical Field, Industrial Field, or both — or whatever the agreement's allocation categories are.

Why the disclosure form matters years later. Inventorship disputes are decided on evidence of who conceived what and when, and the evidence is laboratory notebooks, emails, and disclosure forms created contemporaneously. A collaboration with disciplined disclosure records has a defensible inventorship position; one without has a fact question.

A correction mechanism. Inventorship errors can be corrected, but the process is easier before issuance than after, and easier without deceptive intent — which is why the committee should catch errors during prosecution rather than during litigation.

Funding, milestones, and the commercial terms

The IP allocation is the legal core; the commercial terms determine whether the collaboration happens.

Funding structures:

Each party bears its own costs. Simplest, and appropriate where contributions are roughly balanced and each will exploit the result in its own field.
One party funds the other's work. Common where one party is a customer or a larger company sponsoring a smaller one's development. The funder will expect ownership or an exclusive license, and the funded party will resist. This is the central negotiation.
Shared costs, on a stated split, with a budget and a mechanism for overruns.
Milestone payments, tied to defined technical achievements.

The relationship between funding and ownership. Funders instinctively believe that paying for development means owning the result. That instinct is negotiable and often wrong — a company paying a specialist to improve the specialist's own technology is buying a product, not the technology. The workable compromise is usually exclusivity in the funder's field rather than ownership.

Milestones. Define them technically and objectively: a measurement, a threshold, a test protocol, and who performs the test. A milestone defined as "successful completion of Phase 2" is a dispute. And address what happens if a milestone is not met — a cure period, a renegotiation, a termination right, or a reduction in the other party's rights.

Resource commitments. Named personnel or defined full-time equivalents, for a defined period, with a mechanism for replacement and a consequence for shortfall. A commitment to "use reasonable efforts to devote appropriate resources" commits nothing.

Royalties, where the structure includes them. The base, the rate, the term, stacking provisions where third-party licences are needed, minimum annual amounts, reporting, audit rights, and the treatment of combination products. Combination product royalty definitions are the most litigated provision in technology licensing and deserve careful drafting.

Exclusivity. Whether either party may pursue the same technology alone or with a third party, during the collaboration and after. This is frequently the provision the business people care most about and the one drafted least carefully.

A worked example: the Brentwaite collaboration

The parties. Brentwaite Materials makes specialty polymers. Corvellis Medical makes implantable devices. They agree to develop an antimicrobial coating for Corvellis's implants, using Brentwaite's polymer chemistry.

The term sheet says: "IP developed jointly will be jointly owned."

That sentence is the problem, and counsel's first job is to explain why.

What it produces. Every patent from the project is co-owned. Brentwaite may license the coating to Corvellis's largest competitor, non-exclusively, at any royalty, without consent and without accounting. Corvellis may do the same to Brentwaite's competitors in industrial coatings. And if a third party infringes, neither can sue without the other's cooperation, which — as Ethicon and STC.UNM show — the other can simply withhold.

The restructuring. Counsel proposes a field-of-use allocation:

Corvellis owns foreground IP in the Medical Field, defined as coatings for devices intended for implantation in or contact with the human body, together with associated regulatory classifications
Brentwaite owns foreground IP in all other fields
Each grants the other an exclusive, royalty-free, perpetual license in the fields it does not own
Inventions spanning both fields are owned by Corvellis in the Medical Field and Brentwaite elsewhere, with the patent prosecuted jointly and each bearing its share

The field definition, drafted with care:

"Medical Field" means the use of a Coating on or in any device, article, or material that is (a) intended for implantation into the human body, (b) intended for contact with human tissue, blood, or bodily fluids for a period exceeding [24] hours, or (c) regulated as a medical device by any Regulatory Authority. For the avoidance of doubt, the Medical Field excludes veterinary applications, laboratory and research equipment not intended for human contact, and food-contact applications.

"Industrial Field" means every use other than the Medical Field.

The exclusions matter. Veterinary and food-contact applications are the obvious boundary cases, and the parties resolved them in negotiation rather than in litigation.

Background IP. Schedule 1 lists Brentwaite's polymer patents and know-how; Schedule 2 lists Corvellis's device patents and its regulatory data. Each licenses the other for the project, and — the provision counsel insisted on — Brentwaite grants Corvellis a license to its background polymer IP, in the Medical Field, sufficient to make, use, and sell coated devices, surviving termination.

Why that provision is essential. Without it, Corvellis owns Medical Field foreground IP that it cannot practice, because practicing it requires Brentwaite's underlying polymer chemistry. Owning a patent you cannot practice is owning nothing.

Inventorship management. The agreement establishes a Joint IP Committee meeting monthly, reviewing invention disclosures, determining inventorship with counsel, and deciding on filings. Every disclosure is logged with the contributing personnel and the date.

Why this matters. Inventorship is determined by the claims as allowed, and it changes during prosecution. A committee that reviews the claims before each filing and each amendment catches the moment when an amendment brings in the other party's contribution — which is the moment ownership changes.

Employee assignments. Counsel confirms both parties' employment agreements use present-tense assignment language — "hereby assigns" — and that the seconded personnel are covered. Two Corvellis contractors are found to be working under agreements with an "agrees to assign" clause. They are re-papered before the project starts.

Three years later — the dispute. A Brentwaite engineer develops a curing process that dramatically improves the coating's adhesion. It is used in the medical product. Is it foreground IP in the Medical Field, owned by Corvellis? Or an improvement to Brentwaite's background polymer chemistry, owned by Brentwaite?

The agreement answers it, because counsel addressed the question: improvements to a party's background IP belong to that party, provided that where an improvement is developed in the course of the collaboration and is specific to the Medical Field, it is foreground IP allocated under the general rule. The Joint IP Committee determines the characterization, and disagreements go to an independent expert with relevant technical qualifications, on a 30-day timetable.

The expert determines that the curing process is a general improvement to Brentwaite's polymer chemistry, not Medical Field-specific — it improves adhesion in every application. Brentwaite owns it, and Corvellis has an exclusive license to it in the Medical Field under the background license.

Corvellis is satisfied because it can practice; Brentwaite is satisfied because it owns; and the dispute took five weeks rather than three years.

The counterfactual. Under the term sheet's original sentence, the curing process would have been co-owned, Brentwaite could have licensed it to every one of Corvellis's competitors, and neither could have enforced it against an infringer.

Exit, termination, and survival

Collaborations end, and the agreement should say what happens.

License survival. Do the licenses granted survive termination? For most collaborations they must, because a party that has built a product on licensed technology cannot lose the license because the collaboration ended.

The usual formulation: licenses survive termination other than termination for the licensee's uncured material breach, and even then survive on payment of a stated royalty in some structures.

Prosecution and maintenance. Who continues to file, prosecute, and maintain co-owned or jointly relevant patents, and what happens if one party declines to fund a jurisdiction. The standard mechanism is an offer to the other party, which may take over prosecution and thereafter own that jurisdiction's rights.

Confidentiality and trade secrets. Survive, for a defined period or indefinitely for genuine trade secrets.

Data. Who keeps it, who may use it, and for what.

Regulatory filings. Who owns them, who may reference them, and whether a right of reference survives.

Non-compete and exclusivity. Whether either party may pursue the technology alone, or with a third party, and for how long.

Ongoing obligations. Any royalty or milestone payments, and the reporting and audit rights that support them.

Materials and samples. Return, destruction, or retention.

The provision most often omitted: what happens if one party is acquired by the other's competitor. Collaborations frequently outlive the strategic rationale that created them, and a change-of-control provision — suspension of information rights, a termination right, a buyout of the other's interest — should be negotiated at the outset, when neither party knows which side of it they will be on.

Special contexts

Collaborations with universities. The institution's IP policy governs, and it should be read rather than assumed. Recurring issues: the institution's requirement to publish, which conflicts with patent filing timing and with trade secret protection; its retained right to use the results for research and teaching; government march-in and reporting obligations where federal funding is involved; the institution's reluctance to grant broad exclusive licenses; and the effect of Stanford v. Roche on whether the institution actually owns what its researchers invent. Negotiate a publication review period — 30 to 60 days for the company to review a proposed publication and to request a delay for patent filing — and accept that the institution will publish eventually.

Collaborations with government contractors. Rights in technical data and computer software developed under government funding are governed by acquisition regulations that grant the government defined license rights, and those rights follow the data. A commercial collaboration that touches government-funded development can inadvertently give the government rights in the result, and the analysis should be run before the work starts.

Collaborations with a customer. The customer's standard form usually assigns everything to the customer. For a technology company, signing it is frequently fatal, because it transfers improvements to the company's own core technology. The workable position: the customer owns deliverables and anything specific to its application; the supplier owns its underlying technology and all improvements to it; and the customer receives a broad license to use the deliverables.

Collaborations with a competitor. The antitrust analysis applies: scope the collaboration narrowly, restrict information flows outside the field, and avoid any agreement about price, output, customers, or territories beyond what the collaboration genuinely requires. A joint development agreement between competitors should be reviewed by antitrust counsel before signing.

Collaborations with a startup. The larger party should recognize that its standard form may make the startup unfinanceable — an investor will not fund a company whose core IP is licensed to or owned by a large partner on unfavourable terms. A term sheet that kills the startup's next round has not created a collaboration. And the startup should recognize that the large party's diligence will examine its assignment chain, its open source position, and its freedom to operate.

Cross-border collaborations. Inventorship and ownership rules differ by jurisdiction; some countries require the first filing to be domestic where the invention was made there, or require a foreign filing licence; employee invention compensation is mandatory in several jurisdictions and cannot be contracted away; and export control classification should be run before any technical data crosses a border.

What a diligence reviewer looks for

Joint development agreements are examined in financings, acquisitions, and licensing transactions, and the reviewer's questions are predictable. Drafting with them in mind produces better agreements.

Does the company own what it says it owns?

Present-tense assignment language in employee agreements, not "agrees to assign"
Contractors and consultants under written agreements with present assignments
Academic collaborators' institutional policies examined
Assignments recorded with the patent office
The chain traced from each named inventor to the company

Is anything co-owned? And if so:

Is there an agreement addressing consent, accounting, enforcement, prosecution, and transfer?
Can the company sue alone? If not, the patent is materially less valuable
Can the co-owner license the company's competitors?

Can the company practice what it owns?

Is there a background IP license covering exploitation, not just development?
Does it survive termination?
Is it broad enough in field, territory, and sublicensing?

What has the company given away?

Licenses granted, and their scope
Exclusivity commitments
Non-competes
Grant-back obligations, and whether they are exclusive
Rights that survive termination

What are the field definitions, and do they hold?

Are they precise enough to apply to actual products?
Are the boundary cases addressed?
Is there a mechanism for characterization disputes?

Are there open source or third-party components? In software deliverables, with the licence terms and any copyleft obligations identified.

What happens on a change of control? Termination rights, suspension of information rights, or nothing — which is itself an answer.

The reviewer's summary judgment. A joint development agreement that allocates ownership clearly, grants a surviving background licence for exploitation, addresses co-ownership if any, and has a workable field definition is unremarkable. One that says "jointly developed IP will be jointly owned" and stops there is a diligence finding, and it will be priced.

Quick reference

Co-ownership of a patent is the worst available outcome. Each owner may practice and license without consent and without accounting, and neither can sue without joining the other — a co-owner that refuses to join, or that licenses the defendant, ends the case. Ethicon, Schering, and STC.UNM are the authorities, and they are unforgiving.

Joint inventorship arises easily. Contribution to the conception of one claim confers ownership of the entire patent, and inventorship is determined by the claims as allowed — so it can change during prosecution.

Confirm the company owns its employees' inventions. Present-tense assignment — "hereby assigns" — not "agrees to assign." Stanford v. Roche and FilmTec make the distinction dispositive.

Four allocation structures, in rough order of cleanliness: sole ownership with a license; field-of-use allocation; subject-matter allocation; and co-ownership with an agreement that fixes every default — consent, accounting, enforcement joinder, prosecution, abandonment, and transfer.

Background IP needs three provisions, not one: identification on a schedule; a license to use in the project; and — the one most often missed — a license to exploit the result, surviving termination. A party that owns foreground IP it cannot practice owns nothing.

Copyright differs from patent. Joint authorship is harder to establish, co-owners must account for profits, and either may sue alone.

Trade secrets and data are frequently the most valuable output and the least addressed. Watch the residuals clause.

Manage inventorship actively through a joint IP committee, with disclosure forms capturing who contributed what and when, and counsel flagging every claim amendment that could change the analysis.

Address survival at the outset: licences, prosecution, confidentiality, data, regulatory filings, and — the one always omitted — what happens if one party is acquired by the other's competitor.

And the sentence to strike from every term sheet: "IP developed jointly will be jointly owned."

A note on why these agreements are drafted badly

Joint development agreements are negotiated by business people who are excited about a collaboration, under time pressure, before anyone knows what will be invented. That combination produces the recurring failure.

The specific dynamic. Engineering leadership on both sides wants to start. The commercial terms — funding, milestones, exclusivity — are what the business people care about, and they are settled first. The IP allocation is treated as a legal detail to be papered afterwards. And because nobody yet knows what will be invented, "joint ownership" sounds like a fair and neutral answer rather than what it actually is: an agreement that neither party will be able to enforce anything.

The second dynamic is that the person who understands the technology and the person who understands the legal consequence are rarely in the same conversation. An engineer asked whether an invention would be "an improvement to our polymer chemistry or a new medical coating" gives a clear answer; the same engineer is not asked, because the question is posed years later in a dispute rather than at drafting.

What actually helps.

Draft the allocation before the term sheet is signed, not after. It takes one conversation with the technical leads and it prevents the term sheet from containing a sentence counsel then has to unwind.

Ask the engineers the right question early: what are we likely to invent, and does it belong more naturally to your technology or to theirs? The answer usually produces the allocation structure immediately.

Build the background IP schedule during diligence, when both parties are already disclosing.

Explain the co-ownership default in concrete terms. "Your partner can license this to your largest competitor tomorrow, keep the money, and refuse to help you sue an infringer" changes minds in a way that a citation to section 262 does not.

And accept that the allocation will feel arbitrary at the time. It is being made before anyone knows what will be invented, which is uncomfortable and is exactly why it must be made — because after the invention exists, the parties' positions are fixed by what it turned out to be, and agreement becomes far harder.