As medical device companies navigate an increasingly complex regulatory landscape, the intersection of intellectual property strategy and FDA compliance has become a critical—and often misunderstood—challenge.
Recent high-stakes patent litigation underscores the risks:Siemens Healthineers and Hologic are
currently locked in a multimillion-dollar dispute
over mammography patents, while
Masimo and Apple have battled over pulse oximetry technology
in cases that resulted in import bans and hundreds of millions in damages.
Meanwhile, the Federal Trade Commission continues its crackdown on unsubstantiated health claims, and FDA has intensified scrutiny of promotional materials that exceed authorized indications.
Against this backdrop, device manufacturers face a delicate balancing act: building robust patent portfolios that protect innovation while ensuring marketing claims remain anchored to FDA-cleared indications.
The consequences of getting it wrong extend beyond regulatory warning letters—they can include unenforceable patents, as demonstrated in the landmark Belcher v. Hospira case, where inconsistent statements to FDA and the USPTO rendered patents invalid for inequitable conduct.
Jay Halt, an attorney with intellectual property boutique Volpe Koenig, connected with
MD+DI
to explain how companies should coordinate their IP and regulatory strategies, the biggest mistakes he's seeing in 2026, and why the shift from hardware to data-driven devices demands an entirely new approach to intellectual property protection.
FDA has been cracking down on companies making unsubstantiated claims. From an IP perspective, how should medical device companies coordinate their patent claims with what they are actually allowed to market under FDA clearance? Where do you see companies getting this wrong?
Halt:
Patent claims and marketing claims serve entirely different functions. A patent defines an exclusionary right—it's a "fence" around your intellectual property. It doesn't give you permission to market the invention or establish that it's safe, effective, or FDA-authorized.
There's no requirement that patent scope match FDA-authorized labeling. A well-designed patent portfolio may legitimately cover a technology platform or methods not yet cleared. What you promote, however, must remain anchored to the intended use and conditions supported by your FDA clearance, De Novo grant, or PMA approval. The key is consistency. Patent filings, FDA submissions, quality records, and marketing materials should all tell the same basic story about the product.
Where companies get this wrong: Treating patents as substantiation. "Patented" doesn't establish clinical validity. Claims like "more accurate" or "diagnoses" still require regulatory status and scientific support.
Borrowing broad patent language for advertising. A patent may describe numerous uses, but promoting those uses can create evidence of broader intended use than FDA reviewed.
Making unqualified superiority statements. Arguments made to distinguish prior art for the Patent Office shouldn't become claims of superior clinical safety or effectiveness without proper evidence.
Saying different things to FDA versus USPTO. You can't characterize a feature as conventional in an FDA submission while arguing it's groundbreaking to the Patent Office. The Federal Circuit's Belcher v. Hospira decision illustrates the consequences: the patent was held unenforceable for inequitable conduct.
Operating in silos. Patent, regulatory, quality, and marketing teams must collaborate. By the time an inconsistency is discovered, it may already be embedded in public records. The preferred approach is early coordination among legal and client teams across all regulatory processes to catch problems before they become major issues.
What is your recommended strategy for timing patent filings relative to FDA submissions?
Halt:
My general recommendation is to file when the invention is sufficiently developed to be described and enabled, not when the FDA filing calendar happens to begin. Patent filings should also be done as early as possible, in order to predate possible patent submissions of competitors. In most cases, the first meaningful patent filing should precede the 510(k), De Novo, PMA, or IDE submission. It should also precede any public disclosure of any type, including presentations, publications, clinical-trial disclosures, trade-show demonstrations, investor announcements, unrestricted physician evaluations, offers for sale, and other potentially public disclosures.
A provisional patent application can be an efficient first step because it establishes an early effective filing date and provides a 12-month period in which to file a corresponding nonprovisional application. But a provisional application does not provide any protection. The later claims receive the provisional filing date only to the extent that the provisional supports them. A few slides, a marketing summary, or a conceptual sketch may create the illusion of protection while leaving the commercially important claims unsupported.
I generally recommend the following sequence for patent application filings: First, file a robust initial application. It should describe the core device architecture, alternative embodiments, software and data flows, user interfaces, cybersecurity mechanisms, clinical workflows, and contemplated indications in sufficient technical detail.
Second, conduct invention-harvesting reviews throughout development. Design verification, human-factors work, clinical testing, cybersecurity assessments, and FDA feedback frequently generate additional inventions. A Pre-Submission meeting may change the test plan or device design. An FDA request for additional information may prompt a new algorithm, sensor arrangement, risk control, or manufacturing process. Those developments should trigger supplemental filings where appropriate rather than being forced into an application that did not originally describe them.
Third, file the nonprovisional application before the earliest provisional one-year deadline. Public disclosure before filing can also jeopardize foreign patent rights because many jurisdictions do not provide the same inventor grace period available under U.S. law.
Fourth, maintain a thoughtful continuation strategy. For important platform technologies, keeping a continuation pending can allow claim scope to evolve as the FDA-authorized configuration, competitive landscape, and commercial use become clearer.
How do you protect innovations in AI-enabled medical devices that continue to learn and update after FDA authorization?
Halt:
For adaptive AI, protect the architecture of change, not just a static model. The IP portfolio should be layered. Patent the durable technical architecture. A well-developed specification is key to robust protection. Claims may address data acquisition, model training and retraining, drift detection, acceptance criteria, bias controls, human oversight, cybersecurity, and how the output changes clinical or operational workflow. Consider system, method, and computer-readable-medium claims where appropriate.
Avoid result-oriented claims. Claims like "using AI to improve diagnosis" may encounter patent-eligibility, written-description, enablement, and prior-art issues. Describe the technical mechanisms that produce the result. Use trade secrets for non-public elements. Trade secrets can provide distinct competitive advantages. Candidates include model weights, curated training sets, annotation rules, negative experimental results, and discarded approaches—especially when difficult to reverse engineer.
Use copyright and design protection where appropriate. Copyright can protect source code and documentation. Design patents may cover graphical user interfaces, icons, and displays.
Use contracts to secure commercial implementation. Agreements with hospitals, researchers, contractors, and vendors should address invention assignment, data use, model-training rights, derivative models, confidentiality, open-source components, and rights following termination or acquisition.
How can FDA inspections expose gaps between patent filings and what is actually being manufactured?
Halt:
An FDA inspection is not a patent audit, and a patent does not become invalid merely because the commercial product uses only one of several embodiments disclosed or claimed. Broad patent coverage frequently and legitimately extends beyond the current commercial configuration.
What an inspection can expose is a factual or documentary inconsistency. An investigator may review design and development records, the medical-device file, manufacturing specifications, supplier controls, production records, software versions, verification and validation results, change controls, complaints, nonconformances, cybersecurity documentation, and distribution records. Those materials can reveal, for example:
A patent declaration characterized a component, algorithm, material, or dimensional range as critical, but the released product omits it or treats it as unimportant. A patent application relies on particular performance data, while the underlying quality records contain contrary results or limitations that were not communicated to the patent team.
The FDA-authorized product uses a locked algorithm, while production records show that the commercial version has been updating outside the authorized configuration. A manufacturing process described as essential to obtaining a claimed result was replaced without appropriate documentation, validation, or regulatory assessment. The FDA submission describes one device configuration, while the current bill of materials, software build, labeling, or supplier specification reflects another. Patent counsel was never informed of prior versions, competitor products, negative testing, or publications that may be material to patentability.
Regulatory personnel were never informed of patent-prosecution arguments concerning performance limitations, intended use, or technological differences that may be material to FDA’s review. Again, the mere fact that manufacturing changed does not establish wrongdoing. Innovation requires iteration. The red flags are undocumented changes, inadequate validation, lack of traceability, and conflicting factual representations made to different institutions.
As device value shifts from hardware to data, how should companies protect ownership of real-world evidence?
Halt:
Begin by recognizing that real-world data and real-world evidence are different assets. FDA defines real-world data as information on patient health status routinely collected from sources like EHRs, claims, and registries. Real-world evidence is clinical evidence about a product's use, benefits, or risks derived from analyzing that data. There's rarely a universal answer to "who owns the data." Device manufacturers interact with patients, hospitals, physicians, research organizations, cloud providers, and analytics vendors—each may possess different legal or contractual rights. Treat real-world evidence as a "rights stack" and separately define:
Source medical records and protected health information
Raw device-generated signals and measurements
Metadata, timestamps, and audit trails
Clinical annotations and adjudications
De-identified and limited datasets
Curated, cleaned, normalized, or linked databases
Algorithms, trained models, and model weights
Statistical analyses, reports, publications, and regulatory submissions
Complete contracts before data flows. Expressly allocate rights to collect, access, use, combine, de-identify, validate, retain, submit to regulators, publish, train models, create derivatives, commercialize, sublicense, transfer in acquisitions, and continue using data after the relationship ends. Address security, breach response, audit trails, data deletion, publication review, and restrictions on re-identification.
IP protection layers:
Copyright ordinarily doesn't protect raw facts, but may protect sufficiently original database arrangements—though protection is often thin.
Trade secrets can be powerful for nonpublic datasets, curation methods, annotation protocols, and analytics, provided the company maintains secrecy and the information derives value from not being generally known.
Patents may protect novel methods for collecting, processing, analyzing, or acting on data, but not raw patient facts.
Privacy rights are separate from IP ownership. HIPAA protects identifiable health information and provides de-identification mechanisms.
Most importantly, secure both economic and regulatory-use rights. A dataset may have commercial value yet be unsuitable for FDA use if you cannot verify provenance, inspect source records, evaluate bias, reproduce analysis, retain data, or permit FDA audits.
What are the biggest IP planning mistakes you are seeing device companies make in 2026?
Halt:
The largest mistake is using a static IP plan in a constantly moving regulatory environment. In a short period, QMSR became effective, FDA finalized AI PCCP and RWE guidance, revised General Wellness policy, and USPTO revised AI-assisted inventorship guidance. A 2026 strategy must be a living document.
Recurring mistakes:
1. Keeping IP, regulatory, quality, clinical, and marketing teams siloed. This produces inconsistent statements, missed inventions, undisclosed prior art, and promotional claims that outrun authorization. USPTO has made clear that practices preventing patent personnel from learning material FDA information are not a safe harbor.
2. Filing thin provisional applications and failing to update portfolios as R&D continues. A provisional without commercial architecture, alternatives, and implementation details provides little priority support. Equally problematic is failing to file on improvements generated during design transfer, validation, FDA interactions, and post-market surveillance.
3. Protecting the enclosure but not the platform. Many portfolios remain hardware-heavy even when value has migrated to algorithms, data pipelines, calibration methods, interoperability, cybersecurity, clinical workflow, and user interfaces.
4. Failing to make deliberate patent-versus-trade-secret decisions. Some companies publish through patents what would be more valuable as confidential know-how. Others try keeping readily reverse-engineered features secret, allowing competitors to lawfully replicate them.
5. Neglecting chain of title and third-party restrictions. Common gaps: contractors who never assigned inventions, hospital agreements prohibiting commercial model training, university rights, cloud/AI vendors reserving rights in improvements, incompatible open-source licenses, and datasets whose licenses prohibit regulatory submission. For AI-assisted inventions, weak records of human conception create inventorship problems.
6. Treating FDA authorization as freedom to operate. FDA doesn't determine patent infringement. A device may be substantially equivalent to a predicate and still infringe patents covering the predicate, components, methods of use, software, connectivity, or manufacturing.
7. Measuring success by patent count rather than business coverage. Ten patents on a discarded prototype may be worth less than one family covering the commercial platform, consumables, software-update architecture, and clinical workflow. Portfolio review should ask which revenue stream, market barrier, or acquisition value each asset supports.
Is there anything else you would like to expand on?
Halt:
I recommend a recurring IP-regulatory concordance review at five lifecycle gates:
Concept and platform architecture
Design freeze and transfer to manufacturing
Before the principal FDA submission
Before authorization and commercial launch
Before any material postmarket change or new indication
At each step, the company should bring internal teams together and review patentability, inventorship, ownership, trade-secret status, freedom to operate, data rights, regulatory representations, quality documentation, labeling, and planned promotional claims. The most durable principle is simple: the patent story, the FDA story, the quality story, and the commercial story should be four accurate views of the same product. They need not use identical language or have identical scope, but they cannot contradict one another.
