Artificial intelligence is creating difficulties for the medical device industry operating in the traditional, hardware-driven, highly regulated patent protection environment in which product iterations can move slowly.
Conventional patent strategies for these companies rely on a milestone approach for patent filings—after hardware features are validated, approved, or ready for market. The current patent landscape, with new software features only a few coding sessions away from reality, requires a more speculative approach to software and AI patents.
Integrating AI features into traditional hardware-based medical technologies creates complications for patent protection. However, medical device companies can embrace some strategies to resolve these tensions.
Capturing Future Milestones
In the AI context, patent protection often needs to anticipate functionality that’s still in development. AI innovations, even for medical devices, are often developed independently of regulatory and commercialization timelines, allowing them to be more quickly tested, refined, and incorporated into product releases. AI roadmaps should be used to identify and extrapolate AI-driven innovations early in the patent lifecycle.
Existing product roadmaps should be adapted to include code-based innovations as important milestones, include both hardware and software milestones, and extend far into future for the product timeline.
Harvesting AI inventions also should include anticipated features from predicted next-generation improvements to AI models. One way to do this is scheduling AI-focused brainstorming sessions earlier in the harvesting process. These sessions can include cross-functional teams between engineers, product managers, and counsel and, if scheduled early enough, can surface early-stage AI features.
Prioritizing AI features
Once they’ve been forecast, adding the right milestones to the roadmap means prioritizing AI features. The “right” milestones will be different for each company. Not all AI features need to be patented, and each company should focus on what provides a technical contribution or contributes value to the business.
In determining patent strategies, software companies typically prioritize AI features based on market impact, technical feasibility, and some technology forecasting. While hardware patents often focus on physical design improvements or sensor configurations, AI patents can capture patentable contributions in code rather than circuitry or physical structure.
Patentable AI contributions reside in functions such as how hardware makes decisions, adapts to and processes data, or transforms clinical workflows via earlier or improved notifications.
Being aware of these areas of improvement can lead to disclosures covering AI-driven decision-making methods, improved data processing, customer and patient adaptation, and integration of software and hardware aspects.
Examples of questions teams should ask when prioritizing AI features include:
- Does this feature improve clinical performance or usability in a technically novel way?
- Would this feature be difficult for a competitor to replicate in a similar product or service?
- Is the feature expected to drive differentiation from competitor products or be a core part of the product or service?
If the answer to any of these questions is yes, that feature likely deserves its own milestone on the roadmap—and a plan for patent protection alongside its development.
Using AI Roadmaps
Once charted, companies can use their product roadmaps to develop a filing strategy in a way that anticipates technology and feature evolution.
It’s important that companies don’t repeat the mistakes they made in the last software-based technology wave during the rise of the internet. AI claim drafting presents similar challenges to those faced in that era.
During the internet boom, many companies filed patents with overly broad or vague functional claims (such as “our existing product or service, but on the internet”), which led to unsound claims vulnerable to US Patent and Trademark Office litigation and court challenges.
The same kind of mistake is happening now, with claims invoking AI without clearly describing how it contributes technically to performance, decision-making, or device functionality.
When a future AI milestone appears on a roadmap, the features should be scoped to reflect the technical contribution of that feature, such as model architecture, data processing techniques, or hardware integration and not just the result it enables. One example:
- Weak: A system that claims an AI model to more accurately and quickly detect arrhythmias.
- Better: A system comprising a trained neural network that classifies ECG waveforms by dynamically adjusting feature extraction parameters based on historical patient-specific data.
Companies also can use the roadmap to guide filing strategy. Forecast features can be included in first filed patent applications to lay the groundwork for future continuation applications. Specifically, the roadmap should be used to draft initial applications that contain a sufficient range of features, variations, and anticipated evolutions.
Each meaningful feature milestone on the AI roadmap—whether it’s a new AI capability or a potential new software feature for a hardware product—is a patent protection opportunity. The roadmap isn’t just a guide for when to file, but provides guideposts for identifying features to seek protection, which features are primed for future developments, and keeping strategy aligned with future product value.
This article does not necessarily reflect the opinion of Bloomberg Industry Group, Inc., the publisher of Bloomberg Law and Bloomberg Tax, or its owners.
Author Information
Dohm Chankong is a director in Sterne Kessler’s electronics practice group.
Richard Coller is a director in Sterne Kessler’s mechanical and design and trial and appellate practice groups.
Write for Us: Author Guidelines
