Who owns the IP in a joint development in Japan?

It is common for a startup to develop a new product or service jointly with a large corporate, a university or another startup. One side brings the technology; the other brings funding, facilities, a customer base or data, and together they can build something neither could build alone. One of the hardest questions in practice is then who owns the intellectual property the collaboration produces.

The short version

• "We built it together, so we co-own it" is not a sufficient answer.
• Recent Japanese government guidelines treat ownership, licence-back rights, the scope of exclusivity, consideration, improvement inventions and the effect on an exit as separate design questions.
• The new IP transactions guideline issued in June 2026 goes further, naming "joint research in name only" and disadvantageous treatment of co-owned IP as specific problems.

The instinctive answer is that if the work was joint, splitting the patent 50/50 must be fair. Yet the government guidelines and model contracts published in recent years move away from that "co-own it for now" reflex, and treat ownership, rights of use, field of use, consideration and the duration of exclusivity as elements to be designed separately.

On 24 June 2026 the Japan Fair Trade Commission, the Small and Medium Enterprise Agency and the Japan Patent Office published a new Guideline on Abuse of Superior Bargaining Position and Related Conduct for the Appropriate Transaction of Intellectual Property Rights, Know-How and Data (the "IP transactions guideline"). It covers know-how and data as well as patents, and addresses one-sided allocation of rights in joint development, "joint research in name only", restrictions on using the results, most-favoured-treatment conditions and disadvantageous treatment of co-owned IP.

This article follows the line of guidelines issued by METI, MEXT, the JPO and the JFTC, and considers how intellectual property should be structured when a startup enters joint research or joint development in Japan.

1. The rules have moved from "who owns the IP" to "how the commercial value is shared"

Government documents on joint research and development go back some way.

On industry–academia collaboration, MEXT and METI issued Guidelines for Strengthening Joint Research through Industry–Academia–Government Collaboration in 2016, with a supplement in 2020. They position universities and companies as equal partners and treat university knowledge as value rather than merely a cost.

On collaboration between startups and established companies, the JPO publishes Model Contracts for Promoting Open Innovation (the "OI model contracts") covering NDAs, PoC agreements, joint development agreements and licences. The current version is ver2.2, revised in April 2025, with editions for advanced materials, for AI and for universities.

The JFTC and METI, in the Guideline on Business Collaboration with, and Investment in, Startups, set out the problems that arise in practice at the NDA, PoC, joint research and licensing stages and how the Antimonopoly Act applies to them. The current version was amended on 19 February 2026.

Then, in June 2026, the IP transactions guideline described above was issued. It is not confined to startups: it covers transactions in intellectual property, know-how and data generally, and it comes with a model joint development agreement annexed.

What this sequence shows is that the IP question in joint development cannot be settled as a formal question of title — whether the deliverable belongs to party A, to party B, or to both.

Current practice requires the contract to be designed around who contributes what, who creates what, who uses the results in which market, and how the resulting commercial value is shared.

2. "Joint development" and "joint invention" are not the same thing

The first distinction to draw is between what the contract calls joint research or joint development and what the Patent Act treats as a joint invention.

Signing a joint research agreement does not make an invention a joint invention where only one party substantially contributed to completing it.

The IP transactions guideline proceeds on the basis that a joint inventor is someone who substantially cooperated in bringing the invention into being. It states clearly that, in a case of "joint research in name only", bearing the research costs does not in itself justify acquiring the intellectual property in the results.

Take a project in which the startup holds the core technology and does nearly all of the development work, while the large corporate provides research funding and a test environment. The fact that the corporate bears the cost does not automatically make it reasonable to say that "this is joint development, so every patent is co-owned 50/50."

The guideline's position is that bearing joint development costs does not by itself justify the allocation of the resulting intellectual property, and that where a party who is not an inventor acquires the rights, separate consideration reflecting the contribution to the results is normally required.

Development funding and the price of acquiring intellectual property are, properly, two different things.

3. Start with the background IP

Joint development agreements tend to draw attention to the ownership of results that do not yet exist. For a startup, what matters at least as much is establishing what technology and know-how it already holds going in.

Suppose the startup already has algorithms, source code, trained models, manufacturing methods, experimental data, drawings, customer data and technical know-how.

Once those are used inside the collaboration, the boundary between what existed beforehand and what the collaboration produced can become unclear.

The IP transactions guideline treats this as "contamination risk": technology the startup could have held alone ends up being treated as a result of the joint research, and so becomes a co-owned patent or the counterparty's patent.

It therefore suggests exchanging lists of background information at the outset, and evidencing prior possession of process know-how, data and source code with timestamps and similar means before the contract is signed.

On the startup side, background IP, foreground IP (what the collaboration generates) and improvements arising afterwards should all be separable. Above all, the core technology must not be swallowed by the contract's definition of "results".

4. Why "co-own it for now" can work against a startup

Where neither side is willing to give up the results, co-ownership is often proposed as the fair answer. On its face it is.

For a startup, however, co-ownership of intellectual property is not a neutral outcome.

Under Article 73 of the Patent Act, each co-owner of a patent may work the invention itself unless the contract provides otherwise. But granting a non-exclusive licence to a third party, or transferring a co-ownership share, requires the consent of the other co-owner as a rule.

For a corporate that works the invention in its own factory, co-ownership may cause little difficulty. Startups, by contrast, tend to grow by outsourcing manufacture, licensing to widen the market, moving into adjacent industries, partnering with companies abroad, or selling a business together with its IP.

If each of those steps needs the co-owner's consent, the corporate holds what amounts to a veto. The OI model contracts make the same point, explaining how a co-owned patent can obstruct outsourced or volume manufacture, licensing, and M&A.

So rather than starting from "joint work, therefore joint ownership", work backwards from the question of whose ownership makes the technology easiest to commercialise.

5. Separate ownership from use

The key move in the recent OI model contracts and in the startup guideline is to separate the ownership of the intellectual property from the right to use it commercially.

For example: the patent is owned solely by the startup, while the corporate receives an exclusive right of use confined to a defined scope — "the automotive field, for five years".

The startup collaboration guideline notes that treating a joint invention as co-owned without proper consideration means the startup needs the counterparty's consent every time it wants to use the IP in another application or another field, which impedes its freedom to develop the business.

It offers, as one solution, allocating the intellectual property to the startup and granting the collaborating company an exclusive right of use limited by field and by period.

The result can be sliced: the startup owns the rights; company A has exclusivity in the automotive sector for five years; the startup is free in the medical sector; manufacturing generally is non-exclusive; overseas markets are subject to discussion between the parties.

There is no need to resolve "who owns the rights" and "who may use them where" as a single question.

6. If you grant exclusivity, cut it by scope and by period

The corporate side has a legitimate interest too. Having funded the research and contributed customer data and facilities, it has little incentive to collaborate if the same technology is supplied to its competitors as soon as the work is finished.

Some exclusivity is therefore not the problem. What matters is narrowing it: which market, which territory, which competitors, for how many years, and prohibiting what.

Compare "for ten years after termination, A shall not supply any third party other than B with services using the results" against "for two years after commercial launch, A shall not supply services using the results to the direct competitors of B listed in the schedule, in the [x] industry in Japan". The effect on the startup is not remotely the same.

The 2026 IP transactions guideline takes the same view: restrictions on customers that are reasonable in period and scope, and necessary to preserve the confidentiality of the results, are in principle unobjectionable, while restrictions beyond a reasonable period and scope may amount to abuse of a superior bargaining position or, depending on the facts, to exclusive dealing or restrictive dealing.

7. Watch the clause that hands over every improvement

Joint development agreements often contain a clause providing that inventions arising from improvements to the results also belong to the counterparty.

For a startup this carries enormous weight: a modest improvement to the current service can pull years of subsequent technology into the collaboration partner's hands.

The JFTC's Guidelines on Joint Research and Development under the Antimonopoly Act treat an obligation to grant other participants a non-exclusive licence to improvement inventions as in principle unobjectionable, but classify an obligation to assign improvement inventions to the counterparty, or to license them exclusively, as highly likely to constitute an unfair trade practice, because it blunts the incentive to carry out research and development.

So "improvement invention" should not be defined broadly and casually. Distinguish who made the improvement, whether it arose during or after the joint research period, whether the counterparty's confidential information was used, how closely it relates to the joint research theme, and whether the licence is exclusive or non-exclusive.

8. What it means that the 2026 guideline names "joint research in name only"

The striking feature of the 2026 IP transactions guideline is how concretely it describes actual transactions.

One example: a project labelled joint development in which one party in fact supplied nearly all of the know-how and technology, and was nonetheless asked by the counterparty to co-own the patents 50/50.

Another: a PoC applying an existing patent, in which one party effectively produced the whole deliverable, and the counterparty demanded assignment of all rights on the basis that it had provided the funding.

The message is that what counts is not the label on the contract but who actually supplied what and who actually created what.

Startups in particular tend to reason that receiving development funding from a large corporate makes handing over the IP unavoidable. But payment for development work, payment for acquiring the intellectual property itself, and payment for exclusivity can each be considered separately. Separating out the consideration in this way is central to a joint development negotiation.

9. The 2026 guideline also looks at unfairness after co-ownership

The guideline contains a further point of interest: disadvantageous treatment of co-owned intellectual property.

A patent may be co-owned 50/50 on paper, yet if only one party actually works it in its business while the other is not permitted to, the parties are not substantively equal.

The guideline addresses the case where the results of joint research become a co-owned patent but only one party works it while the other cannot, and indicates that unilaterally demanding a restriction on working may raise an abuse of superior bargaining position issue.

In other words, a 50:50 split of the shares does not make the arrangement fair. What matters is who may work the invention, who may license it, who may commercialise it and who receives the economic benefit. When reviewing a joint development agreement, look past the ownership percentages to whether the rights can actually be used.

10. In AI and software, the subject matter of "IP" is itself more complex

In AI and software, the value produced by a collaboration is rarely confined to patents.

In a generative-AI or machine-learning service there are raw data, pre-processed data, training datasets, source code, algorithms, prompts, models, model parameters, trained models, evaluation methods, experimental results, data obtained in operation, and know-how acquired during development.

That the 2026 IP transactions guideline expressly covers know-how and data, not only patents and copyright, reflects that environment. "Data" there extends to experimental data from research and development, industrial data, CAD data, customer data and sales and marketing data.

In AI there is a further question: whether exclusivity is rational at all. The JFTC/METI startup guideline points out that in an AI business which receives data from several companies, builds customised models and supplies them to several companies, giving each company exclusivity over its deliverable can be irrational for the corporate as well as for the startup.

Going non-exclusive allows the startup to deploy the technology more widely and so reduce the cost of the service, while companies that contributed to the development can be compensated through preferential usage fees. Again, the design question is not whether ownership is transferred but which economic benefits are conferred and how.

11. A joint development agreement affects fundraising and M&A

A startup cannot look only at the moment of signing. The agreement has to be read against future financing rounds, partnerships, licences, business transfers and M&A.

If a core patent is co-owned with a large corporate and the transfer of the share requires that corporate's consent, the point will surface in M&A due diligence.

If an important technology is subject to a permanent prohibition on licensing to competitors, the field of potential acquirers may be narrowed.

When the JPO revised the OI model contracts to ver2.2 in April 2025, one of the stated reasons was to review clauses and commentary that could obstruct the M&A of deep-tech startups, so that M&A becomes a more available option.

A joint development agreement, in short, is not merely a research project contract. It bears directly on the startup's future enterprise value.

12. Under competition law, separate the collaboration itself from the restrictions attached to it

The JFTC's long-standing Guidelines on Joint Research and Development under the Antimonopoly Act matter here as well. They are not new: adopted in 1993 and last revised in 2017.

They start from the proposition that joint research and development is generally pro-competitive — it reduces development cost, spreads risk, shortens timelines and allows technologies from different fields to complement each other.

Where the parties compete with each other, however, a collaboration can reduce research competition itself. As a rough benchmark, joint research on product improvements or substitute products by parties competing in a product market is generally not problematic under the Antimonopoly Act where their combined market share is 20% or less.

More importantly, the collaboration may be unobjectionable while the terms attached to it are not.

The guidelines treat some restrictions as reasonable in certain circumstances — restricting research with third parties on the same theme during the joint research period, or prohibiting use of confidential information outside the agreed purpose to the extent necessary.

By contrast, prohibiting research on themes unrelated to the collaboration, broadly prohibiting research on the same theme after it ends, prohibiting a party from using its own existing technology, prohibiting research that uses the results, requiring improvement inventions to be assigned or exclusively licensed to the counterparty, and restricting the resale price of products using the results are all identified as likely to raise problems under the Act.

So in a joint development agreement, look beyond the IP ownership clause to the non-compete, exclusivity, customer restrictions, improvement technology and sales terms around it.

13. What a startup should settle in a joint development agreement

Taking the above together, at least the following should be worked through before signing.

The point is to negotiate these individually. A single sentence providing that "intellectual property shall be co-owned by the parties" designs almost nothing about the commercial relationship that follows.

14. In conclusion: the contract designs a future business, not a division of IP

The government guidelines of recent years have made the thinking on joint research and development contracts reasonably clear.

The key points: ① bearing the research costs is not the same as acquiring the intellectual property; ② calling something joint research does not make it a joint invention or produce co-ownership; ③ existing technology and the results of the collaboration must be clearly distinguished; ④ "co-own it for now" is not necessarily fair; ⑤ ownership and use should be designed separately; ⑥ exclusivity should be limited by field, territory and period; ⑦ improvement inventions and post-collaboration business activity should not be over-constrained.

It is telling that the 2026 IP transactions guideline addresses not only one-sided allocation of rights but also joint research in name only, restrictions on using the results, most-favoured-treatment conditions and disadvantageous treatment of co-owned IP.

When a joint development agreement is signed, no one knows whether results will emerge at all. Deciding in advance, with any precision, who should own them is difficult.

That is exactly why the contract should set the rule beforehand: which contribution earns which rights and which economic benefits.

For a startup, intellectual property does not exist only for the current product. It is an asset that makes up enterprise value in the next business line, the next financing round, expansion into other industries, and at an exit.

Making a collaboration work therefore calls for more than protecting IP. The contract should be designed around how the intellectual property the collaboration produces will be deployed in practice, and how both parties can share in its value.

Principal sources

• JFTC, SME Agency and JPO, Guideline on Abuse of Superior Bargaining Position and Related Conduct for the Appropriate Transaction of Intellectual Property Rights, Know-How and Data (24 June 2026)
• JFTC and METI, Guideline on Business Collaboration with, and Investment in, Startups (31 March 2022, amended 19 February 2026)
• JFTC, Guidelines on Joint Research and Development under the Antimonopoly Act
• JPO, Model Contracts for Promoting Open Innovation ver2.2
• MEXT and METI, Guidelines for Strengthening Joint Research through Industry–Academia–Government Collaboration (supplement)

Ministry guidelines on intellectual property are collected in the guideline index for intellectual property; to search across fields, use the index filterable by industry and function. These documents are published in Japanese.

---

This article is general information, not legal advice on any specific matter. Whether particular terms are appropriate depends on the technology, the parties' contributions and the commercial plan. For review or drafting of a joint research or joint development agreement under Japanese law, please use the contact form.