This is the third in our series leading up to HLTH USA 2026, where our team will share insights on the legal and business issues facing physician innovators, healthcare leaders, entrepreneurs, and investors.

At HLTH USA 2026, the most useful question to ask may not be “What can your technology do?” It may be “What can your company actually protect?”

As I prepare to head to Las Vegas for HLTH USA 2026, that is the question I would encourage medtech founders to bring to their partnering and investor meetings. The scale of AI-enabled medical device adoption deserves attention. The FDA reports more than 1,600 AI-enabled medical devices authorized for U.S. marketing as of September 2026, with examples ranging from image enhancement and diabetic retinopathy detection to automated insulin dosing.[1] Beyond devices, AI is becoming part of the everyday clinical workflow. In the AMA’s 2026 physician survey, 72% of respondents reported incorporating at least one named AI use case, compared with 48% in 2024 and 38% in 2023.[2] Those are self-reported results, and the survey’s use-case list has evolved, but the direction is hard to miss.

For founders, that expansion changes the IP conversation. A compelling demo is only the starting point. The harder questions sit underneath it: Who contributed to the invention? What rights came with the data and model? Which parts of the product can competitors copy, and which parts have a defensible layer of protection? A useful pre-HLTH exercise is to answer those questions before someone across the table asks them.

AI-Assisted Innovation. Human-Driven Patent Strategy.

Using AI in R&D does not automatically take patent protection off the table. But the inventor still needs to be human.

The USPTO’s November 2025 revised guidance rescinded its earlier AI inventorship guidance and confirmed that the ordinary inventorship standards apply.[3]

At bottom, AI is a tool, not a co-inventor. The question remains whether the human inventor conceived the claimed invention. For a medtech team, that makes contemporaneous development records worth the effort. Record who developed the particular technical solution, how it evolved, and how AI was used along the way. Keep records of the engineering decisions, not just the final output. A prompt history may help explain the process, but it cannot substitute for evidence of human conception.

The next question is what the patent would actually cover. “We use AI to improve diagnosis” is a business description, not a patent strategy. Last year, the Federal Circuit rejected claims that applied generic machine learning in new settings without a claimed technological improvement.[4] That was not a medtech case, and it does not mean AI inventions are categorically unpatentable. What it does mean is that applying generic machine learning to a new field, without more, does not supply the inventive concept required by the court’s eligibility analysis.

Work with patent counsel early and throughout your development process to identify the specific technical advance that enables the technology. Some examples might include a new way to process noisy sensor signals, a model architecture that improves device operation, or an implementation that reduces memory demands. Then explain how it works.

The disclosure must support the scope you want to claim; a desired clinical result is not a substitute for an enabling technical description. Focus on advances that matter to the product and that a commercially useful claim could actually reach.

You Built It. Do You Own the Rights?

Inventorship and ownership are fundamentally different questions. Identifying the right human inventors does not finish the company’s ownership analysis. Review the assignments from employees, consultants, and collaborators rather than assuming that paying for development settles the issue.

What does the company own, what does it license, and what can it continue using if the relationship ends? In a pilot with a hospital or development partner, spell out in your development agreement whether the company may use the data only to perform the pilot, or also to train, fine-tune, validate, and commercialize a product. Address the resulting models and improvements, permitted reuse, and post-termination rights. Do not leave all of that to a broad sentence saying one party “owns the data.” Privacy permissions, confidentiality obligations, and commercial-use rights need separate attention.

AI-generated code and content raise a different ownership trap. A vendor’s promise that “you own the output” does not itself create copyright. Under U.S. law, copyright requires human authorship; protectable human selection, arrangement, or modification may be covered even when generated material is not. The Copyright Office’s registration guidance draws that distinction.[5] Preserve records of human-written code and creative edits, and review assignments, third-party licenses, and vendor terms rather than assuming the entire deliverable comes with exclusive rights.

Keep Your Competitive Edge Out of the Prompt

Not every valuable asset belongs in a patent application. Consider whether nonpublic training methods, curated datasets, tuning techniques, or validation know-how are better candidates for trade-secret protection. Under 18 U.S.C. § 1839, in order to qualify as a trade secret, the information must have economic value derived from its secrecy, and the owner must take reasonable measures to preserve it. Marking a folder “confidential” is a starting point, but is not enough on its own.

Before uploading source code, unpublished designs, or development data, check the actual service terms and settings applicable to the platform you are uploading to. Look specifically for provisions that permit the provider to retain inputs, use them for training, allow human review, or share them with subcontractors.

Where the information warrants it, require contractual use restrictions, limited access, and defined retention and deletion practices. Controlled sharing with a service provider need not destroy secrecy, but the protections should match the information’s value.

AI Can Create a Brand. It Cannot Clear It.

AI can also be a useful starting point for a name or logo. However, using AI to research trademark availability is not a substitute for a clearance opinion performed by a trademark attorney. Before using an AI generated design on booth signage, a product interface, or your website, have trademark counsel assess the name and design against existing marks for the same or related goods and services. A good clearance search will extend beyond exact matches and federal registrations to relevant marketplace use. An AI tool’s assurance that a design is original is not a substitute for that review.

The trademark question should be treated separately from a copyright question. An AI-generated logo may lack copyright protection in its generated artwork, yet still function as a trademark if it identifies and distinguishes the source of the company’s goods or services and otherwise meets trademark requirements. Conversely, copyright protection for a designer’s human contribution does not clear the logo for trademark use. The two rights protect different things: copyright protects creative expression, while trademark is a source identification. Finally, if using an AI tool to design a logo or create a trademark, review the tool’s commercial-use terms, document any human creative contributions, and obtain appropriate rights from outside designers.

The following strategy works whether using an AI tool or not: (i) choose a name that distinguishes the business rather than merely describes the technology; (ii) adding “AI” to a descriptive product phrase is not, by itself, a distinctiveness strategy; (iii) suggestive, arbitrary, and coined marks generally offer a stronger starting point; (iv) descriptive marks face protection and registration hurdles; and (v) generic terms cannot function as trademarks.

Be careful to keep approved names and logos consistent across AI-assisted marketing, and review generated copy and imagery for misleading suggestions of affiliation, endorsement, or sponsorship.

Before You Pitch, Get Your IP Story Straight

Before the conference, consider the following: what, if anything, needs filing with the USPTO before a public presentation, give your team a clear disclosure playbook, and confirm any assignments and licenses that support the company’s IP story. Finally, remember to keep your public deck useful enough to start a conversation without giving away what makes the technology difficult to reproduce.

Not every conversation at HLTH should turn into an IP diligence session. However, it is important to be able to explain the specific steps you’ve taken to make the next conversation easier. Show what the product can do, then be ready to explain what the company controls, what it can protect, and where it depends on someone else’s permission. AI may accelerate development. A thoughtful IP strategy helps make sure the value created along the way stays with the business.

I’ll be joining my Husch Blackwell colleagues at HLTH USA 2026 in Las Vegas, where we’ll be presenting as part of the Physician Innovator Series on November 15. We’ll be discussing these issues and other legal and business considerations facing physician entrepreneurs, healthcare leaders, and investors.

We’ll share additional perspectives in this series leading up to the conference.

If you’re attending HLTH, I’d welcome the opportunity to connect, hear what you’re working on, and discuss some of the challenges and opportunities you’re seeing.

[1] FDA, Artificial Intelligence-Enabled Medical Devices, “List of AI-Enabled Medical Devices” (reporting more than 1,600 authorizations as of September 2026), https://www.fda.gov/medical-devices/digital-health-center-excellence/artificial-intelligence-enabled-medical-devices.

[2] American Medical Association, 2026 Physician Survey on Augmented Intelligence (Mar. 2026), “Physician Use of AI,” pp. 5–6, https://www.ama-assn.org/system/files/physician-ai-sentiment-report.pdf/. The chart distinguishes incorporation of named AI use cases (72%, 48%, and 38%) from uncertainty about available tools. Results are self-reported; the use-case list expanded in 2026.

[3] USPTO, Revised Inventorship Guidance for AI-Assisted Inventions, 90 Fed. Reg. 54,636, 54,636–37 (Nov. 28, 2025), §§ II–IV, https://www.federalregister.gov/documents/2025/11/28/2025-21457/revised-inventorship-guidance-for-ai-assisted-inventions. Section II rescinds the February 2024 guidance; §§ III–IV address human conception and AI’s role as a tool.

[4] See Recentive Analytics, Inc. v. Fox Corp., 134 F.4th 1205 (Fed. Cir. 2025).