Your NDA Protects Your Secrets.  Your Patents Protect Your Inventions.  But Who Protects the Opportunity You Created?

By Daniel J. Holmander and Michel “Mike” Morency, Ph.D. on October 28, 2026

Why AI Is Changing the Economics of Life Sciences Innovation, and Why Patents and NDAs Alone May No Longer Be Enough

The Short Version An NDA protects only what it defines as confidential, and only if that information really is confidential.  In a 2026 case, the Federal Circuit wiped out more than $17 million in trade secret damages, largely because what the plaintiff shared under an NDA had already appeared in decades-old patents. Patents cover what you claim and can support.  They don’t cover every other route to the same clinical result, and AI is making those routes easier to find. Identifying an opportunity is not inventing the solution.  A potential partner whose scientists conceive the solution will usually own it. Contracts can fill the gap.  A well-drafted contract, signed before the important disclosures, can capture value that patent and trade secret law leave on the table. Investors should treat issued patents and signed NDAs as the beginning of diligence, not the end of it.

The Partnership That Never Happened

Picture an early-stage life sciences company.  Its founders have spent years on an unmet clinical need.  They’ve brought together researchers and clinicians, worked out the science, and developed an approach that could genuinely help patients.  Maybe it’s a drug delivery platform, a minimally invasive surgical tool, a better formulation, or an AI-enabled diagnostic.  They have preliminary data, prototypes, and pending patent applications.

What they don’t have is the money or infrastructure to finish clinical development, get through the FDA, manufacture at scale, and sell the product.  So, they go looking for a strategic partner.

A large pharmaceutical or device company agrees to take a meeting.  Both sides sign an NDA.  The startup walks through the clinical problem, its solution, the expected benefits, and the size of the market.  It shares early results but holds back formulation parameters, manufacturing details, and other specifics.  The larger company is interested, asks good questions, and schedules more meetings.  Then it passes.

About a year later, it announces its own program aimed at the same problem.  Its scientists have come up with a different formulation, a different molecule, or a different device mechanism, possibly with help from AI tools for modeling and optimization.

Let’s assume the larger company did nothing wrong.  It didn’t misuse anything confidential.  Its scientists did their own work, and its product doesn’t fall within the startup’s patent claims.  Still, the startup is the one that found the opportunity, showed it was worth pursuing commercially, and put it in front of the larger company.  The larger company may be entirely free to go after it anyway.

That raises the central question for life sciences innovators:  Who protects the value of spotting the opportunity when someone else can lawfully build the solution?

AI Is Shrinking the Innovator’s Head Start

Life sciences has always been a hard field to break into.  Developing a drug takes specialized expertise, years of lab work, preclinical and clinical studies, and a great deal of capital.  Devices require engineering, prototyping, validation, clinical evaluation, manufacturing, and regulatory clearance.  Those hurdles have historically given innovators time to establish themselves before anyone could catch up.

AI is lowering some of them.  Researchers now use it to identify targets, predict molecular properties, screen candidates, analyze biological data, model protein structures, and explore alternative device designs.  It doesn’t replace experiments, clinical trials, or regulatory review, and biology is still unpredictable.  What it does well is speed up the early, exploratory work:  finding approaches worth testing and weeding out the ones that aren’t.

That’s a problem for innovators, because so much of their value lies in knowing where to look.  A biotech company may spend years establishing that a particular pathway is worth targeting.  A device startup may be the first to see that a new surgical approach could solve a clinical problem.  Once that insight is out, a competitor with good AI tools can explore alternatives faster and more cheaply than it could have even a few years ago.  It may still have years of development ahead, but one of the hardest parts, figuring out which direction to go, has been done for it.

The worry isn’t only copying.  It’s that competitors can increasingly build something scientifically different that goes after the same clinical opportunity and do it entirely within the law.

What a 2026 Federal Circuit Case Teaches About NDAs

International Medical Devices, Inc. v. Cornell, Nos. 2025-1580, -1605 (Fed. Cir. Apr. 17, 2026), is a useful, if sobering, example.

International Medical Devices (IMD) makes the Penuma® cosmetic penile implant.  In 2018, a urologist attended a Penuma surgical training session.  He signed an NDA and watched implant procedures.  During the session, the implant’s inventor, Dr. James Elist, talked about three ideas for improving the device:  internal cavities to make it softer, mesh tabs near the tip to encourage tissue ingrowth, and absorbable sutures to hold the tabs in place.  The urologist later received IMD’s list of surgical instruments and supplies.  He went on to help develop a competing implant, and patents issued on aspects of it without naming Dr. Elist as an inventor.

IMD sued for trade secret misappropriation, breach of the NDA, and other claims, and it won at trial.  The jury found that all four asserted trade secrets (the three design ideas plus the instrument list) had been misappropriated, that the NDA had been breached, and that the competitor’s patents were invalid for leaving out Dr. Elist.  The district court awarded roughly $5.8 million in reasonable royalties, $11.5 million in exemplary damages, and $1 million in statutory damages for trademark counterfeiting, and it entered a five-year injunction.

The Federal Circuit reversed most of that.  Applying California’s Uniform Trade Secrets Act, the court held that none of the four items was a protectable trade secret.  The three design ideas had been described in patents from 1980 and 1992 or were otherwise long known in the field.  It didn’t matter that those earlier designs had never been commercialized, because a patent disclosure makes information “generally known” either way.  Nor did it help that IMD was applying known concepts to a cosmetic implant rather than a therapeutic one.  As for the instrument list, IMD had emailed it to the urologist, and to a third party, without marking it confidential or taking other steps to protect it.

The NDA claim failed for the same reason.  The agreement excluded publicly available information, and IMD couldn’t point to any confidential information beyond the four alleged trade secrets.  The court vacated the royalty and exemplary damages and the injunction.  Only the trademark counterfeiting award survived.

Then came the part that should worry any innovator.  The court also reversed the inventorship ruling and reinstated the competitor’s patents.  Dr. Elist’s ideas were already known, so they didn’t count as contributions to the conception of the claimed inventions.  And the claims had been allowed only after the applicants added a hardness limitation that Dr. Elist had nothing to do with.

A few practical points come out of the case:

• Signing an NDA doesn’t make everything you say confidential.  What’s protected depends on how the agreement defines confidential information, what it carves out, and whether the information is actually non-public.
• Check the prior art before calling something a trade secret.  If an idea, or something close to it, already shows up in a patent or a journal article, it probably isn’t a secret.
• Treat confidential material as confidential.  Mark it, limit who gets it, and keep a record of what you disclosed, to whom, and when.
• The other side may end up holding the patents.  An idea the law doesn’t protect can be developed by someone else into an invention they own.

Patents Cover Your Invention, Not Every Way to Get There

A valid patent lets its owner stop others from making, using, selling, offering to sell, or importing the claimed invention.  Unlike a trade secret plaintiff, a patent owner doesn’t have to prove copying, and independent development, with or without AI, is generally no defense.  When research is moving this fast, that’s a real advantage.  But patents have two limits that executives should understand.

The first is scope.  Life sciences patents have to meet strict enablement and written-description requirements, and courts have been skeptical of broad claims that define a molecule by what it does rather than what it is.

• In Amgen Inc. v. Sanofi, 598 U.S. 594 (2023), the Supreme Court unanimously struck down claims to antibodies defined by their ability to bind a particular region of the PCSK9 protein and block its interaction with LDL receptors.  If you claim an entire class of compositions, the Court said, your patent has to teach skilled artisans how to make and use the entire class.  Amgen’s “roadmap” and screening methods amounted to “little more than two research assignments.”
• In Ariad Pharmaceuticals, Inc. v. Eli Lilly & Co., 598 F.3d 1336 (Fed. Cir. 2010) (en banc), the Federal Circuit confirmed that written description is a separate requirement from enablement.  The patent has to show that the inventor actually possessed what is claimed.
• In Juno Therapeutics, Inc. v. Kite Pharma, Inc., 10 F.4th 1330 (Fed. Cir. 2021), the Federal Circuit threw out a judgment of about $1.2 billion and held broad CAR-T claims invalid for lack of written description.  The patent didn’t adequately describe which binding domains would work against which targets.

Diagnostics face an extra obstacle.  Claims directed to a newly discovered natural correlation are vulnerable to patent-eligibility challenges under Mayo Collaborative Services v. Prometheus Laboratories, Inc., 566 U.S. 66 (2012).

In practice, this means that discovering a biological mechanism matters doesn’t give you rights to every molecule that might someday act on it.  A device company can’t claim every mechanism that produces the clinical result it’s after.  A diagnostics company usually can’t lock up a natural biological relationship just by identifying it.  That leaves room for competitors to develop commercially important alternatives without infringing, and AI is making those alternatives easier to find.

The second limit is timing.  Under the safe harbor in 35 U.S.C. § 271(e)(1), activities reasonably related to developing information for submission to the FDA generally don’t infringe, even if they use the patented invention.  Merck KGaA v. Integra Lifesciences I, Ltd., 545 U.S. 193 (2005).  So, a competitor can often run its preclinical and clinical programs while your patent is in force.  The patent really starts to bite at commercialization.

Patents are still enormously valuable.  They just may not capture everything that’s valuable about having found an important scientific or clinical opportunity.

Identifying the Opportunity Is Not Inventing the Solution

Inventorship depends on conception.  In Burroughs Wellcome Co. v. Barr Laboratories, Inc., 40 F.3d 1223 (Fed. Cir. 1994), part of the AZT litigation, the Federal Circuit described conception as “a specific, settled idea, a particular solution to the problem at hand, not just a general goal or research plan.”  The court largely rejected the argument that NIH scientists who did important confirmatory testing should have been named as joint inventors.  To be a joint inventor, a person has to contribute significantly to the conception of at least one claim.  Pannu v. Iolab Corp., 155 F.3d 1344 (Fed. Cir. 1998).  AI doesn’t change any of this.  Revised USPTO guidance issued in November 2025 treats AI as a tool, not an inventor, and applies the same conception standard to AI-assisted inventions.

That can lead to an uncomfortable result in partnering talks.  Suppose a startup identifies a promising clinical application and convinces a larger company it’s worth pursuing.  If the larger company’s scientists then conceive a specific composition, formulation, or device design, the startup’s people generally aren’t inventors of it, even though they pointed the way.  Identifying the problem, spotting the commercial opportunity, making introductions, or suggesting a research direction can all be valuable.  None of them makes you an inventor or gives you a claim to the patent.  IMD v. Cornell shows exactly this:  the competitor kept its patents.

The time to sort out who owns future inventions is before the scientists start working together.

Contract Law Can Fill the Gap

Patent and trade secret law aren’t the only options.  A contract can require a company to pay for using an idea that was disclosed to it, even if no patent ever issues.

The classic example is Aronson v. Quick Point Pencil Co., 440 U.S. 257 (1979).  An inventor licensed a keyholder design to a manufacturer for a 5% royalty, which would drop to 2.5% if no patent issued within five years.  The patent application was rejected.  The manufacturer argued that the agreement was unenforceable, but the Supreme Court disagreed, holding that federal patent law didn’t preempt the royalty obligation.  The manufacturer had agreed to pay, and the contract didn’t take anything out of the public domain or stop anyone else from copying the design.

Courts applying New York law have reached similar results in idea-submission cases.  Under Apfel v. Prudential-Bache Securities Inc., 81 N.Y.2d 470 (1993), and Nadel v. Play-By-Play Toys & Novelties, Inc., 208 F.3d 368 (2d Cir. 2000), a contract to pay for a disclosed idea can be enforced if the idea was novel to the buyer, even if it wasn’t new to the world.  Other states handle this differently, so the governing-law clause deserves attention.

For life sciences companies, this means a partner’s payment obligations don’t have to hinge entirely on whether a patent ends up covering the final product.  There are limits, of course.  The agreement has to reflect a real, sufficiently definite bargain.  And none of these cases lets a company claim an entire therapeutic area just because it raised the subject first.

A Better Way to Structure the Partnership

The usual sequence is to sign a mutual NDA and start presenting.  That puts most of the disclosure up front and leaves the economics for later.  We think it works better to proceed in stages, matching each round of disclosure to a corresponding commitment from the other side.

Stage 1: Initial confidential evaluation.  Share enough to make the case: the clinical need, the scientific rationale, the supporting data, and the commercial potential.  Be selective about sensitive technical details.  Read the other side’s NDA form carefully.  Big-company forms often include “residuals” clauses (which allow use of information retained in unaided memory), broad carve-outs for independent development, or obligations that limit disclosure without clearly limiting use.  Ask that confidential information be used only to evaluate the specific transaction, and make sure oral and visual disclosures, such as presentations, site visits, and procedure observations, are covered too.

Stage 2: Paid evaluation or option.  If the partner wants to dig deeper, negotiate the terms:  how long the evaluation lasts, what it can access, and what it pays.  That might be a fee for an exclusive negotiation period or an option to acquire specified development and commercialization rights.

Stage 3: Definitive agreement before joint development.  Before the scientists start working together in earnest, decide who owns, and who can license, background IP, improvements, jointly developed inventions, research data, and resulting products.  In life sciences, the agreement should specifically address new analogs, alternative formulations, improved delivery systems, manufacturing processes, and other follow-on technology.

Plan, too, for the possibility that the deal doesn’t happen.  Can the partner use research results generated during the collaboration?  Can it keep working in the same therapeutic area?  What has to stay confidential?  Are rights in jointly developed technology kept, licensed, or terminated?  These questions are much easier to answer before anyone has created value than after the parties have gone their separate ways.

Leaving Room for Legitimate Independent Work

A good agreement doesn’t have to stop the larger company from doing its own research, and usually it shouldn’t try.  Large pharmaceutical and device companies often have several programs running on related targets, indications, or platforms.  A blanket ban on working in a therapeutic area will probably be a non-starter in negotiations, and it may not be enforceable anyway, whether as an unreasonable restraint of trade under state law or under the antitrust laws.  In California, for example, restraints in business-to-business contracts are judged under a rule of reason.  Ixchel Pharma, LLC v. Biogen, Inc., 9 Cal. 5th 1130 (2020).

A better approach is to spell out exactly what information and rights the startup is providing and what the recipient agrees to give in return.  For especially sensitive deals, the parties can document existing research programs, limit which employees see particular confidential materials (a “clean team”), and keep records that separate independent research from work that draws on protected information.

AI makes all of this more important.  The agreement should say whether confidential information can be put into AI tools, whether internal models or third-party platforms, who can see the outputs, and how AI-generated results that draw on confidential inputs will be treated.  Nobody is trying to stop scientific progress.  The point is to have clear rules about which research is allowed and which isn’t.

Rethinking Patent Strategy

Contracts should work alongside a strong patent strategy, not replace it.  Before a major disclosure, it’s worth asking whether more development work could support broader or more commercially useful claims.  For example:

• Pharmaceuticals: alternative compounds, formulations, dosage forms, delivery systems, and methods of treatment.
• Medical devices: alternative mechanical architectures, component arrangements, control mechanisms, and procedural methods.
• Diagnostics: specific technical improvements to testing platforms, sample processing, detection systems, and analytical methods.

Whatever comes out of that work has to be a real invention: conceived, adequately described, and enabled.  Having an AI tool generate a list of hypothetical alternatives doesn’t make any of them your invention and publishing that kind of list can create prior art against your own later applications.  Used carefully, though, AI can help you think through how a well-funded competitor might design around your core technology, and that exercise can turn up patentable improvements before a partner or competitor finds them.  It can also help you decide what to patent and what to keep as a trade secret.

What you want is a portfolio built around the alternatives that matter commercially, not a stack of narrow patents that are easy to design around.

What Investors and Acquirers Should Ask

For venture capital, private equity, and strategic investors, all of this goes to valuation.  Heavy R&D spending doesn’t guarantee a lasting competitive advantage, and issued patents don’t, on their own, prove that a company’s core opportunity is protected.  Good diligence questions include:

• Would the patent claims cover the alternatives a well-funded competitor using AI-assisted design is most likely to pursue?
• What has the company disclosed in partnering discussions, to whom, and under what agreements?  Do any of those NDAs have residuals clauses or other gaps?
• Do the company’s collaboration agreements protect its rights in improvements and follow-on developments?
• Are the claimed trade secrets actually secret, or are they sitting in the patent and scientific literature?

It works the other way, too.  An acquirer should find out whether the target’s own programs carry obligations to universities, inventors, former collaborators, research institutions, or other companies whose ideas or information contributed to them, whether through licenses, material transfer agreements, sponsored research agreements, or confidentiality agreements.  Those obligations can affect freedom to operate, development plans, and the economics of bringing a product to market.

The Bottom Line

Life sciences runs on shared ideas.  Universities partner with biotech companies, startups work with pharma, physicians collaborate with device developers, and investors fund all of it.  That isn’t going to change, and it shouldn’t.  What is changing is the risk of sharing what you know, because AI makes it cheaper for others to act on it.

Often a startup’s biggest contribution is seeing something first: a promising therapeutic pathway, an unmet clinical need, or proof that a technology can solve a real medical problem.  Patents protect inventions that are properly disclosed and claimed.  Trade secret law protects information that is actually kept secret.  Neither one guarantees that the innovator will share in the value of every independently developed solution to the problem it identified.

Closing that gap takes a coordinated plan.  Build a patent portfolio around the alternatives that matter commercially.  Protect real know-how.  Control what you disclose and when.  And negotiate evaluation fees, options, or license terms before you give anyone access to something with real economic value.  A company that learns where to aim its research has gained something valuable.  You can put a price on that, but only before you give it away.

Your NDA protects your secrets.  Your patents protect your inventions.  But protecting the opportunity you created may require a commercial agreement that recognizes its value before someone else turns that opportunity into the next successful therapy, diagnostic, or medical device.

Selected Legal Authorities

NDA and trade secrets

1. International Medical Devices, Inc. v. Cornell, Nos. 2025-1580, -1605 (Fed. Cir. Apr. 17, 2026).

Patent scope, eligibility, and timing

• Amgen Inc. v. Sanofi, 598 U.S. 594 (2023).
• Ariad Pharmaceuticals, Inc. v. Eli Lilly & Co., 598 F.3d 1336 (Fed. Cir. 2010) (en banc).
• Juno Therapeutics, Inc. v. Kite Pharma, Inc., 10 F.4th 1330 (Fed. Cir. 2021).
• Mayo Collaborative Services v. Prometheus Laboratories, Inc., 566 U.S. 66 (2012).
• Merck KGaA v. Integra Lifesciences I, Ltd., 545 U.S. 193 (2005); 35 U.S.C. § 271(e)(1).

Conception and inventorship

• Burroughs Wellcome Co. v. Barr Laboratories, Inc., 40 F.3d 1223 (Fed. Cir. 1994).
• Pannu v. Iolab Corp., 155 F.3d 1344 (Fed. Cir. 1998).
• USPTO, Revised Inventorship Guidance for AI-Assisted Inventions (Nov. 2025).

Contract protection for disclosed ideas

1. Aronson v. Quick Point Pencil Co., 440 U.S. 257 (1979).
2. Apfel v. Prudential-Bache Securities Inc., 81 N.Y.2d 470 (1993).
3. Nadel v. Play-By-Play Toys & Novelties, Inc., 208 F.3d 368 (2d Cir. 2000).
4. Ixchel Pharma, LLC v. Biogen, Inc., 9 Cal. 5th 1130 (2020).

This article is for general informational purposes only and does not constitute legal advice.  Protection for disclosed scientific ideas and commercial opportunities depends on the specific agreement, governing law, and circumstances.