The invention clause we wrote too late — the 6 steps I now run before any engineer signs in Japan

The conversation I want to help you avoid goes like this. An engineer who joined eighteen months ago has built something genuinely patentable. Someone in the business asks whether it is protected. Someone in legal pulls the employment contract, finds a general confidentiality clause and nothing about inventions, and discovers that under Japanese law the right to obtain that patent belongs to the engineer personally.

Nobody did anything wrong. The contract was a competent contract. It just was not a Japanese invention-assignment contract, and in Japan that is a specific instrument with a specific timing requirement that cannot be satisfied afterwards.

This guide covers the six steps we now run before an engineer signs. It is written for companies building engineering teams in Japan, particularly those doing it without a local legal function. As always: this is operational guidance from practice, not legal advice, and the clause itself should be drafted by a Japanese patent attorney (弁理士) or employment lawyer.

The rule in one paragraph

Article 35 of the Japanese Patent Act governs inventions made by employees. Its default is the opposite of what most non-Japanese employers assume: the right to obtain a patent arises in the inventor, a natural person, not in the company that paid for the work. The employer’s position depends entirely on what was agreed before the invention happened.

The amendment effective 1 April 2016 added what is now the decisive provision. Where a contract, work rules or other stipulation provides in advance that the right to obtain a patent belongs to the employer, that right vests in the employer from the time the invention occurs — original acquisition, with nothing to transfer and no moment at which the engineer owned it. Absent such a provision, the employee owns it and the employer holds only a statutory non-exclusive licence.

Step 1: Test whether it is a shokumu hatsumei at all

Article 35 does not cover everything an employee invents. Three conditions must hold together:

1. The invention falls within the scope of the employer’s business. Read broadly in practice, but not infinitely — an engineer at a fintech who patents a bicycle mechanism is outside it.
2. The act of making it belongs to the employee’s present or past duties. Note past: an invention flowing from work the engineer did in a previous role at your company still qualifies.
3. The inventor is an employee of the company (the provision also covers officers, and national and local government employees).

Where all three hold, you are in the Article 35 regime and the advance-provision question decides ownership. Where they do not — a genuinely unrelated side invention — it is a free invention, and a clause purporting to capture it is, to that extent, ineffective. Blanket clauses claiming everything an engineer invents during employment are common in templates imported from other jurisdictions and are one of the reasons those templates travel badly.

Step 2: Put the provision in place before the invention exists

This is the step that carries all the weight, and the one that cannot be repaired later.

The advance provision can live in an individual employment contract, in the work rules (就業規則), or in a separate invention-handling regulation referenced by them. What matters is that it exists before the invention occurs, and that it is validly part of the employment terms for the engineer in question.

Two failure modes we see repeatedly:

• The rules exist but were never made known. Work rules operate on the basis of having been properly established and communicated. An invention regulation living in a drive nobody has been shown is a weak foundation for an ownership claim over a valuable patent.
• The rules exist but not for this person. An engineer who joined through an acquisition, transferred from an overseas entity, or was engaged first as a contractor and converted, may sit outside the document set the clause lives in. Check per person, not per company.

If you are hiring your first engineers in Japan, the invention regulation belongs in the same pass as the work rules themselves — the same pass in which the social insurance and payroll setup gets built. Doing it later means doing it for people who already have inventions.

Step 3: Take original acquisition, not succession

Before 2016 the standard mechanism was succession: the invention arose in the employee, and the employer succeeded to it under a pre-agreed obligation to assign. The 2016 amendment added original acquisition as an alternative, and where the drafting supports it, it is clearly the better route.

The joint-invention row is the one that surprises people. Where a patent is co-owned, a co-owner generally cannot transfer its share without the other co-owners’ consent — so under the succession route, an invention made jointly by your engineer and a partner company’s engineer can leave you needing a third party’s cooperation to acquire your own employee’s share. Original acquisition sidesteps it, because there is no transfer to consent to.

Step 4: Build the reasonable benefit standard through the three-step procedure

Taking the right is one half. The other is what the inventor receives for it, and this is where litigation actually happens.

The 2016 amendment reframed the entitlement from reasonable remuneration to reasonable monetary or other economic benefit (相当の利益). The widening was deliberate: benefit can take forms other than a cash payment, including stock options, promotion, funded study abroad, research sabbatical and paid further education. For a startup that is short on cash and long on equity, that flexibility is the point of the provision.

But the amendment paired the flexibility with process. Article 35(5) frames the question of whether the benefit is unreasonable around three things, and METI has issued guidelines on how to run them:

1. Consultation with employees when establishing the standard by which the benefit is calculated.
2. Disclosure of that standard to the employees it applies to.
3. Hearing the inventor’s views on how the standard is applied to their specific invention.

Read those as a checklist, because that is how they function in a dispute. A moderate benefit produced by a documented consultation, a published standard and a recorded hearing is a far stronger position than a generous benefit handed down unilaterally. Employers who lose these cases usually lose on the absence of process, not on the size of the number.

Practical shape of a standard that works for an engineering team: a fixed filing award on application, a fixed grant award on registration, and a variable component tied to actual exploitation with a stated calculation method and review point. Write down what happens when an invention has several inventors, and what happens when the inventor has left. Both will occur.

Step 5: Remember that software splits across two statutes

This is the step most often missed, and it is specific to engineering teams.

Article 35 is patent law. It says nothing about copyright, and the code your engineers write is primarily a copyright asset. Copyright in Japan is governed by a separate regime with a different default in the employer’s favour.

Under Article 15 of the Copyright Act, a work created by an employee in the course of duties, on the employer’s initiative, is authored by the employer — and for computer programs, Article 15(2) applies without the requirement that the work be made public under the employer’s name that applies to other work types. So for ordinary in-house development, copyright in the code vests in the company by operation of law, subject to no contrary provision in a contract or work rules.

Two consequences worth acting on. First, do not let a well-drafted copyright clause create false confidence about patents — they are different statutes with opposite defaults, and a contract that handles code does not handle inventions. Second, check that your invention regulation does not accidentally reduce your copyright position: a clause saying the company acquires IP “on assignment by the employee” can be read as a contrary provision displacing the Article 15 default, converting an automatic entitlement into one contingent on paperwork.

Step 6: Cover contractors, joint inventors and cross-border teams

Article 35 protects employers of employees. Three groups sit outside it.

Contractors

An engineer engaged on a business entrustment (業務委託) contract is not an employee, so neither original acquisition nor the statutory non-exclusive licence is available. Rights stay with them unless expressly assigned. Your contractor agreement therefore needs an assignment clause doing work that the employment framework does automatically for staff.

The sting: if that contractor is later found to have been an employee in substance, you inherit the Article 35 analysis for a period in which no work rules covered them. The classification question and the IP question are the same question wearing different clothes — our guide to gyomu itaku and worker status covers the substance tests.

Joint inventors

Where an invention has multiple inventors across entities, each entity acquires only its own employees’ share, and co-ownership rules then govern what any one of you can do with it. Agree the position in the collaboration agreement before the work starts, not when the application is being drafted.

Engineers outside Japan

Employee invention rules are national. An engineer employed by your Japanese entity is in Article 35; one employed by a group company elsewhere, or engaged through an employer of record, is governed by that other relationship — and where an EOR is the legal employer, the invention regulation that matters may be theirs rather than yours. Ask to see it. It is a routine request and the answer is occasionally alarming.

Teams running this problem across several jurisdictions at once will find the shape differs sharply by country: HireDeveloper.sg covers the Singapore position, where the default sits closer to the employer, and HireDeveloper.ae the UAE framework. Japan is the outlier of the three in starting from inventor ownership.

A 30-minute audit you can run this week

For an existing team, in order:

1. Does an invention provision exist in the employment contract, the work rules, or a referenced regulation? If none of the three, you are on the default and every existing invention belongs to its inventor.
2. Does it use original acquisition language, or only an obligation to assign?
3. Is there a written benefit standard, and can you evidence consultation, disclosure and a hearing route?
4. Which engineers are outside the document set — acquisitions, transfers, converted contractors, EOR staff?
5. Does anything in your contracts displace the Copyright Act Article 15 default for code?

Where step 1 fails and inventions already exist, the remedy is individual assignment agreements negotiated now, with reasonable benefit, before anyone leaves. Cooperation from a current engineer is generally straightforward. Cooperation from a former one is a different conversation, and occasionally an expensive one.

Frequently asked questions

If we have no invention clause, do we lose the patent entirely?

You lose ownership, not use. The right vests in your engineer personally, and Article 35(1) leaves you a free statutory non-exclusive licence to work the invention in your business. So you can keep shipping. What you cannot do is stop the inventor assigning it elsewhere, grant an exclusive licence, or treat it as an asset in a financing or sale. The licence prevents disaster and contributes nothing to the balance sheet.

Can the clause be added after the invention was made?

Not retroactively. Article 35(3) requires the provision to have existed in advance, so a clause signed in March does not capture a February invention. For those you need an individually negotiated assignment with reasonable benefit — and an engineer who has just learned they own a valuable patent negotiates from a materially different position than one signing a standard clause on day one.

Does reasonable benefit have to be cash?

No. The 2016 amendment widened it precisely to admit non-cash forms: stock options, promotion, funded study abroad, research leave. Process matters more than form — a modest benefit set through proper consultation, disclosure and hearing is more defensible than a generous one imposed unilaterally.

Does Article 35 cover contractors?

No. It governs employer and employee. A business entrustment engineer is outside it entirely, so there is no original acquisition and no statutory licence; rights stay with them absent an express assignment in the contract. And if that contractor is later reclassified as an employee, you acquire the Article 35 problem retroactively for a period with no work rules covering them.

The short version

Japanese invention law starts from a position most foreign employers get backwards: your engineer owns what they invent, unless you wrote something down first. The fix is inexpensive and takes a few weeks — a provision in the work rules using original acquisition language, a benefit standard built through consultation and disclosure, a contractor clause doing the same job by contract, and a check that your copyright position has not been accidentally undermined.

The only genuinely expensive version of this problem is the one discovered late, because the one repair the statute does not permit is doing it retroactively.

This is an operational guide written from hiring practice and is not legal advice. Statutory references are to the Japanese Patent Act and Copyright Act as in force at 14 August 2026. Invention regulations, benefit standards and assignment clauses should be drafted or reviewed by a Japanese patent attorney (弁理士) or employment lawyer before use.