US counsel structuring a life-sciences research collaboration reach for a familiar toolkit: a mutual limitation-of-liability cap, cross-indemnities, and a schedule of required insurance. In a domestic deal these devices are load-bearing. They define the outer edge of what either party can lose. Carry the same toolkit into a collaboration governed by, or performed in, Switzerland, and the edges move. The cap, the indemnity, and the insurance certificate each rest on assumptions that Swiss and EU law do not share, and the divergence surfaces at the one moment the toolkit exists to address: the point of failure. The exposure a US-headquartered sponsor negotiates against is rarely the exposure it actually carries.
1. The Cap a Simple Partnership Does Not Recognize
A limitation-of-liability clause is a term of a contract. It allocates risk between the two parties who signed it. Swiss law, however, does not take the label the parties put on their arrangement as decisive. It qualifies a relationship by its objective substance: where two or more persons combine efforts or resources to pursue a common purpose, the arrangement is a partnership under Art. 530 Abs. 1 OR, and a simple partnership, an einfache Gesellschaft, under Art. 530 Abs. 2 OR unless it meets the conditions of some other company form the statute provides for,1Obligationenrecht (SR 220), Art. 530 ff.: the simple partnership, its loss-sharing, joint assets, and liquidation; Art. 100 Abs. 1: an advance exclusion of liability for unlawful intent or gross negligence is void. whatever the cover page calls it. A co-development agreement, a research-services agreement, a joint steering committee with shared go/no-go authority and a shared stake in the outcome: any of these can carry the objective marks of a common purpose pursued with common means. Whether a particular collaboration is a simple partnership is a question of qualification, and it turns on how the parties actually behaved, not on how they styled the document. A simple partnership can arise by conclusive conduct, and even where neither party set out to form one.
The consequence is that a body of mandatory and default partnership law attaches to a relationship the parties may never have intended to create. Absent a contrary agreement, each partner shares equally in profit and loss regardless of the nature or size of its contribution, under Art. 533 OR. Assets brought into or acquired for the common purpose are held jointly under Art. 544 Abs. 1 OR, and where the partners have incurred an obligation toward a third party, they are jointly and severally liable for it under Art. 544 Abs. 3 OR, subject only to a different agreement with that third party. On dissolution, contributions are returned and any shortfall is borne as a loss, so a partner who expected to risk only its own contribution can be called to absorb a share of the deficit under Art. 549 Abs. 2 OR. None of this machinery reads the limitation-of-liability cap the parties negotiated, because the cap was drafted to bound contractual liability, not to serve as a partnership loss-sharing agreement or to displace a solidary obligation owed to someone who never signed it.
A ceiling drafted to bound what one party owes the other says nothing about what both parties jointly owe a third, and recharacterization is what turns the second exposure on.
Recharacterization also reaches assets the parties never meant to pool. Property and rights transferred to or acquired for the common purpose are held jointly under Art. 544 Abs. 1 OR, and held so in accordance with what the partnership agreement provides, which is where the difficulty sits: whether background patents, materials, or data that an academic institution or a smaller partner contributed to the work were drawn into the joint estate at all turns on what the parties are taken to have agreed about them. Where they were, they fall under the liquidation regime of Art. 548–550 OR, and Art. 548 OR leaves a partner who contributed a thing in ownership with a claim to the value at which it was taken over, not to the return of the thing. The relationship also carries a duty of loyalty among the partners that a purely contractual counterparty does not owe. And because the simple partnership has no legal personality, it supplies none of the entity-level liability shield the parties may have assumed they were operating behind. Each of these is a legal consequence of the qualification, not a term either side can neutralize by drafting the cap more firmly.
The bridge for a US reader is a risk it already knows: a course of dealing that a court reads as a partnership or de facto joint venture by conduct, notwithstanding a disclaimer. The Swiss refinement is sharper than that. A negotiated cap can still bind the partners as between themselves, because the internal profit-and-loss split of Art. 533 OR is a default the parties are free to vary, though that freedom has a ceiling of its own: an agreement made in advance to exclude liability for unlawful intent or gross negligence is void under Art. 100 Abs. 1 OR, and whether a clause that caps rather than excludes falls inside that prohibition is not answered on the face of the provision. What the cap cannot do is bind a third party who never agreed to it: against that party each partner answers for the whole under Art. 544 Abs. 3 OR, and dissolving the partnership does not shed the obligation, which survives the winding-up under Art. 551 OR. The device the parties trusted to bound their exposure governs only the internal split, at exactly the moment an outside claimant, an injured trial subject, a misled counterparty, a regulator, is the one who has appeared. The apparent answer, that the cap bounds the downside, holds right up to the moment the collaboration is recharacterized, and it is the collapse of the collaboration that most often triggers the recharacterization. The shape of this problem for R&D partnerships generally is examined in Insight 01; the concern here is narrower, the specific containment devices US counsel rely on and the specific ways each fails.
2. The Trial-Injury Cover That Does Not Switch Off
Where the collaboration includes a clinical trial conducted in Switzerland, a second containment device fails in a different way. Liability toward a research subject injured in connection with the trial is not a matter the parties are free to allocate as they please. Under Art. 19 HFG, the party that causes a trial to be conducted bears a liability, causal and not dependent on proof of fault, for damage a participant suffers in connection with the project, and under Art. 20 HFG that liability must be adequately secured, whether by insurance or in an equivalent form.2Humanforschungsgesetz (HFG, SR 810.30), Art. 19 and Art. 20: trial-subject-injury liability and the mandatory securing of it. The implementing ordinance, the KlinV, fixes the coverage: minimum insured sums graduated by the risk category of the trial, and a requirement under Art. 13 Abs. 3 KlinV that the cover extend to damage arising up to twenty years after the trial concludes, a tail the ordinance lengthened from ten years to twenty with effect from 1 November 2024. It also cuts holes in both layers. Art. 10 KlinV releases from the Art. 19 HFG liability whoever proves that the damage is attributable to one of the listed causes, among them the administration of an authorized medicinal product used in accordance with its product information, and Art. 12 KlinV takes out of the securing duty both those exempted damages and category A trials whose measures for taking biological material or collecting health-related personal data carry only minimal risks and burdens.3Verordnung über klinische Versuche (KlinV, SR 810.305): coverage of trial-subject liability, minimum sums by category, a 20-year run-off after the trial, and the exemptions from both liability and the securing duty.
Two features of this regime confound the US assumption. The first is that the liability toward the subject is statutory and subject-facing. A cap or an indemnity negotiated between the sponsor and a contract research organization or an investigator site reallocates exposure as between those parties; it does not diminish what the injured subject may recover, nor does it move the securing obligation off the party the regime places it on. The second is temporal. The obligation to maintain cover is measured from the trial's conclusion, not from the life of the collaboration, and the ordinance does not spell out what an early termination does to it. The text points in two directions at once. Art. 13 Abs. 3 KlinV measures the tail from the conclusion of the trial, while Art. 45 KlinV, imposing a retention duty of the same twenty years, writes Abschluss oder Abbruch, conclusion or discontinuation, in terms. What an abandoned trial does to the securing duty is therefore something to be argued rather than read off the provision. What is clear is narrower: the insurer cannot cancel once a covered event has occurred under Art. 14 KlinV, and who funds the run-off while the relationship is being unwound is precisely the question a termination-for-convenience clause tends not to answer.
US counsel are calibrated for a different baseline. The United States has no federal statute mandating subject-injury insurance for clinical trials; compensation for trial injury is a matter of negotiated contract and institutional policy, and it varies widely. A US party accustomed to treating trial insurance as a schedule item, something to be capped, waived, or allowed to lapse on termination, meets in Switzerland a mandatory, continuing, subject-facing obligation that the contract can describe but cannot switch off. A US-style insurance schedule, built around a US carrier, US policy limits, and a claims-made trigger, is not self-evidently the cover the KlinV demands, which is measured against Swiss minimum sums and a duration that outlasts the policy period a US buyer would ordinarily purchase. The agreement may allocate responsibility for procuring and paying for the cover. It cannot allocate away the statutory consequence of a gap, which under Art. 19 HFG attaches to whoever caused the trial to be conducted. Art. 2 KlinV gives that role to a sponsor, which must itself have a seat or a representative in Switzerland, and reaches the investigator only where the investigator has taken on responsibility for causing the trial to be conducted as well. Which entity in a US-Swiss structure answers to that description is a question of how the trial was in fact set up, not of how the agreement allocated the cost of the cover.
3. The Section 365(n) Election That Stops at the Border
US life-sciences counsel treat the insolvency of a licensor as a substantially solved problem, because 11 U.S.C. § 365(n) solves a version of it. Where a debtor-licensor rejects an executory license, the licensee may elect to retain its rights to the intellectual property as they existed immediately before the case commenced, for the term of the contract, in exchange for continuing to pay royalties and giving up the licensor's future affirmative performance.411 U.S.C. § 365(n) and § 101(35A): the licensee's retention election, and an IP definition that omits trademarks and foreign patents; Mission Product Holdings (2019): rejection is a breach and cannot rescind rights already granted. That last point is easy to underweight: the licensee keeps the invention frozen at the petition date and loses the improvements, maintenance, and know-how transfer the collaboration existed to deliver, which for an evolving biologic, cell line, or software platform is much of the value. A collaboration license drafted in the United States is nonetheless often built on the assumption that this election will be available if the counterparty fails.
Two structural limits defeat that assumption in a US-Swiss collaboration. The first is internal to the statute. The definition of "intellectual property" that § 365(n) protects, in 11 U.S.C. § 101(35A), reaches inventions and patents protected under title 35 of the US Code, which is to say US patents, and it does not list trademarks. A European or Swiss patent is not protected under title 35, so the portion of a licensee's rights that runs on foreign patents sits outside the election, as do any trademark rights. Sitting outside the election is not the same as being extinguished. In 2019 the Supreme Court held in Mission Product Holdings that a rejection under § 365 has the same effect as a breach of the contract outside bankruptcy and so cannot rescind rights the contract had already granted, which leaves the holder of foreign-patent or trademark rights arguing from ordinary contract principles instead of a statutory election, and arguing them wherever the underlying rights sit. The second limit is jurisdictional and more fundamental. Section 365(n) is a mechanism of the US Bankruptcy Code, triggered by rejection in a US bankruptcy case. The insolvency of a Swiss counterparty proceeds under Swiss insolvency law, which supplies no codified licensee-election of this kind. The nearest Swiss analogue is the general rule on continuing contractual relationships in bankruptcy, Art. 211a SchKG,5SchKG (SR 281.1), Art. 211a: the general continuing-contract rule in Swiss bankruptcy, not a license-specific election. which is not license-specific and whose contours the statute does not delineate.
The protection is also one-directional in a way a mutual collaboration tends to obscure. Section 365(n) runs to the licensee when the licensor is the debtor; it gives a debtor-licensee's counterparty no matching election. In a collaboration where each side licenses intellectual property to the other, only the party that happens to be the non-debtor licensee in a given insolvency draws the benefit, and if the failing entity is the Swiss partner, the US party is looking to Swiss law for a mechanism its own drafting quietly assumed. The error a US drafter is prone to is to picture the Swiss position as a weaker version of § 365(n). It is not a weaker version; it is a different instrument, and for the two scenarios a cross-border collaboration is most likely to face, a foreign-patent portfolio and an insolvency administered abroad, the election the license presupposes may simply not be in play. The insolvency asymmetry between the two systems, and the reliance on the general SchKG regime in place of a codified election, is developed at length in Insight 13; what matters here is that a device US counsel treat as portable protection is, for much of what a US-Swiss deal exposes, not portable at all.
4. Jointly Owned Results, Opposite Default Rules
The fourth failure lives in the results clause, and it is the one most likely to be papered over with a single word. US counsel read "jointly owned" against 35 U.S.C. § 262: absent an agreement to the contrary, each joint owner of a US patent may make, use, offer to sell, or sell the patented invention within the United States, or import it into the United States, and by settled Federal Circuit reading license it to third parties, without the consent of and without accounting to the other owners.635 U.S.C. § 262: each joint owner may make, use, offer to sell, or sell the patented invention within the United States, or import it, without the consent of, or accounting to, the others; Ethicon (Fed Cir 1998): co-owners may also freely license others, and an infringement action must join all co-owners. Joint ownership, in the US default, is close to independent freedom to exploit. Swiss law inverts that default. Under Art. 33 Abs. 2 PatG, a co-owner of a patent may exercise the rights in the jointly owned patent only with the consent of the other co-owners, retaining independently only the ability to dispose of its own share and to sue for infringement.7Patentgesetz (PatG, SR 232.14), Art. 33 Abs. 2: a co-owner may exercise the rights in a jointly owned patent only with the others' consent; Art. 34 Abs. 2: a license may be granted only with the consent of all those entitled. On licensing the statute is express rather than doctrinal: under Art. 34 Abs. 2 PatG, where a patent application or a patent is jointly owned, a license may be granted only with the consent of all those entitled. How much room is left for a co-owner to work the invention for itself is the harder question, argued out against the co-ownership rules of the ZGB rather than answered on the face of Art. 33 Abs. 2 PatG. Many, though not all, European jurisdictions apply the same consent-to-license default.
The two defaults also fail in opposite directions on enforcement. Under US law the freedom cuts both ways, because all co-owners generally must join to sue for infringement: a co-owner that has independently licensed a competitor, which § 262 permits, can neutralize enforcement by declining to join, so "jointly owned" can mean both parties may license and neither may reliably sue. The Swiss consent rule trades that risk for its mirror image, since no co-owner can license unilaterally but any single co-owner can block all third-party licensing by withholding consent. The same two words therefore describe a near-frictionless exploitation right on one side of the Atlantic and a mutual veto on the other.
Layered on top is a competition-law dimension a US reader would not expect to find attached to a property question at all. Where the collaboration is between competitors and the agreement may affect trade between Member States, so that Art. 101(1) TFEU is engaged at all, the safe harbor runs through the R&D Block Exemption, Regulation (EU) 2023/1066. It is available only where the combined market share of the parties does not exceed 25% on the relevant product and technology markets when the agreement is entered into, and only on conditions governing access to and exploitation of the results: full access to the final results, including any resulting intellectual property rights and know-how, and, where the agreement does not provide for joint exploitation of the results, access to any pre-existing know-how of the other parties that is indispensable for exploiting them.8Regulation (EU) 2023/1066 (R&D Block Exemption): access to the final results (Art. 3), access to pre-existing know-how absent joint exploitation (Art. 4), and the 25% competitor market-share threshold (Art. 6). The way the parties allocate and exploit joint results is therefore, at once, a property question under Swiss patent law and a condition of an EU competition-law exemption, and a clause optimized for one may sit awkwardly with the other. The word "jointly" carries all of this, and rarely announces that it does.
5. Strategic Considerations
What unites these four failures is a single mechanism. Each containment device is drafted against a home-law default, and the governing or performance law quietly substitutes a different one. The substitution is invisible while the collaboration is working and becomes load-bearing the instant it is not. The questions that follow are the ones the collaboration's own documents rarely answer. Would the collaboration, as it was actually performed rather than as it was styled, be qualified as a simple partnership, and if so, was the negotiated cap ever meant to double as the loss-sharing and solidary-liability settlement that qualification imposes? If a trial is halted mid-course, who carries the continuing insurance obligation once the relationship is being unwound, and does the contractual allocation of that cost even coincide with the party the statute holds? Does the insolvency protection the license assumes reach the foreign-patent rights and the foreign-administered insolvency the deal is most likely to encounter, or does it stop at the two limits the US drafter did not price? Does "jointly owned" mean the same thing to both parties, and to a competition authority reviewing the same results clause?
None of these questions resolves on the face of the agreement, and each depends on facts, on conduct over the life of the collaboration, and on commercial context that only the parties possess. The uncomfortable feature they share is that the device meant to make the downside legible is the very device that obscures it, because it invites the reader to believe the exposure has been bounded when the bound is the one thing the governing law has not agreed to. Whether a given collaboration has closed that gap or merely papered over it is not a question a limitation-of-liability clause can answer about itself.