IP and data rules differ by jurisdiction and by institution — route any agreement through your research contracts office, and expect your partner to do the same.
Collaboration agreements get signed for the funder. The clauses that matter later — who holds the data, who can reuse it, what happens to anything patentable — are often filled in from a template nobody read closely.
Those clauses only become contentious when something turns out to be worth money or worth a second paper, and by then neither side can discuss them neutrally.
Separate what each side brings from what the project creates
Background IP — methods, software, datasets, materials, cell lines each party already holds. The standard position is that each keeps ownership and grants the other a licence limited to this project.
Specify two things your template may leave open: whether that licence survives the project, and for what purpose. "For research use only" and "for any purpose" are very different grants, and a partner who signs the second without their contracts office seeing it may not be able to honour it.
Foreground IP — what the project generates. Joint ownership sounds equitable and is the most awkward to operate, because many jurisdictions require all co-owners to consent to each licensing decision. If you choose it, write down now whether each party may use the results in their own research without asking, and what a third-party licence requires.
Data: the terms that actually get argued about
Custody. Where the raw data lives and who holds the authoritative copy. If it was collected in Vietnam, national rules may apply to its transfer and storage — ask your partner to confirm with their institution rather than assuming your own framework governs.
Reuse beyond the agreed outputs. This is the clause most often missing and most often disputed. May either side use the same dataset for a further paper, a student's thesis, a follow-on grant? Silence here reads as permission to one side and as a breach to the other.
Open-data obligations. If your funder mandates deposit, say so at the start and settle who decides what is deposited and when. A partner learning at submission that their data must be made public may have made commitments to participants or to their institution that this breaks.
Where human participants are involved, the consent they gave sets the outer limit. No agreement between institutions can authorise a use that participants did not agree to.
Materials and samples
If biological materials, specimens or samples cross a border, a separate transfer agreement is normally required, and permits may apply under national legislation and international frameworks on access to genetic resources.
Three practical points: start early, because permits take months; make sure the agreement covers derivatives and not only the original material; and confirm who may deposit a derived strain, line or sequence in a public repository, since deposit is effectively a transfer to everyone.
Software and code
Increasingly the most valuable foreground output and the one least often addressed.
Agree the licence for project code up front, and check dependencies — some open-source licences require derivative works to carry the same terms, which forecloses commercial options later. For most academic collaborations, choosing a clear open licence at the start is simpler than resolving it afterwards.
And name who maintains the repository after the project ends. Code with no maintainer is a result nobody can reproduce.
Publication and patent timing
If the work could produce something patentable, disclosure before filing can destroy protection in some jurisdictions.
The workable arrangement is a short, fixed review window before submission — commonly measured in weeks — with an explicit maximum. Open-ended review rights are a standing block on your partner's ability to publish, and for an early-career researcher on their side that is a serious cost.
Make the limit explicit, and make clear that silence within the window means consent.
What to ask your partner's side
Ask three questions early, and ask them of the institution rather than only the researcher:
Who has authority to sign an agreement of this kind — the individual researcher usually does not.
Does the institution have an IP policy covering research results, and can you see it.
How long does their internal review typically take.
Expect the third answer to be longer than your own timeline assumes. Build it into the schedule rather than pressing for a signature; a researcher who signs something their institution had not cleared is exposed personally, and the agreement may not hold.
Match the paperwork to the risk
A small collaboration with no funding, no personal data and no commercial prospect does not need a long instrument — a one-page memorandum on authorship and data use is proportionate.
A full agreement is warranted when any of four things is present: substantial funding, human-participant data, plausible commercial output, or biological materials crossing a border.
The general rule holds in both directions: these terms are easy to settle while nobody knows what the results are worth, and hard to settle once someone does.
What's the difference between background and foreground IP?
Background is what each party already holds, normally staying with its owner under a project-limited licence; foreground is what the project generates and needs an explicit ownership term.
Which data clause is most often missing?
Reuse beyond the agreed outputs — whether either side may use the dataset for a further paper, a thesis, or a follow-on grant.
How should publication review before patent filing be handled?
A short fixed window with an explicit maximum, and silence within the window treated as consent — open-ended review rights block your partner's publishing.
Who should sign the agreement on the partner side?
Someone with institutional signing authority, not the individual researcher — and expect their internal review to take longer than your timeline assumes.
