When it comes to protecting artificial intelligence innovations, our Intellectual Property law team at Nixon Peabody often hears the same question: Should we patent it, or keep it as a trade secret? The answer, as you might expect, is nuanced. In our recent webinar, we unpacked the key considerations that go into making this decision, especially in the ever-evolving world of AI.
Watch a replay of our recent “Patents vs. Trade Secrets” webinar:
Why trade secrets still matter
We started by looking at AI trade secret protection, an often overlooked but powerful form of IP protection. Trade secrets are intended to keep valuable information confidential to maintain a competitive edge. There’s no formal registration process, no waiting period, and no need to disclose your invention publicly, which is an advantage for AI developers working in fast-paced environments.
Trade secrets can cover everything from training data and algorithms to abstract ideas and prototype concepts. They can be beneficial when your invention is still evolving or when you’re not ready to share the “secret sauce” with the world.
But there’s a catch. Trade secrets don’t offer exclusivity. If someone else figures out your method—whether through reverse engineering or independent development—you’re out of luck. And in AI, where many teams are racing toward similar solutions, that’s a real risk.
We discussed the Waymo v. Uber trade secrets case as a prime example. Uber settled for $245 million after being accused of misappropriating trade secrets related to autonomous vehicle technology. That case demonstrates how valuable and vulnerable AI-related secrets can be.
The power of patents
Shifting to patent protection, a patent offers a government-backed right to exclude others from using your invention. The exclusory right provided by patent protection provides a compelling advantage, especially if you’re looking to attract investors, build a portfolio, or monetize your technology through licensing.
Utility patents are the most relevant types of patents for AI inventions, as they can protect methods, systems, and devices. But here’s the challenge: patent eligibility. AI inventions often involve abstract ideas, mathematical concepts, or mental processes, all of which are tricky under the Alice/Mayo framework for subject-matter eligibility.
To demonstrate subject-matter eligibility under the Alice/Mayo framework, a claimed invention needs to provide a technical improvement, which may include providing a more efficient computing system, a novel model architecture, or a new, technical way of applying AI within an organization.
Examples of AI-related innovations that have been examined and allowed by USPTO include:
- Use of machine-learning techniques to predict and reduce credit risk based on contextual customer data
- Modular AI pipelines so internal teams can customize workflows
- Providing explainability tools through intuitive user interfaces
These inventions don’t simply automate human tasks; they improve how systems operate, and that distinction is often key to meeting subject-matter eligibility requirements under the Alice/Mayo framework.
Making the call: Patent or trade secret?
Deciding between patent protection and trade secret strategy involves more than assessing patentability. The most effective approach depends on a range of factors, including business objectives, development timelines, and how easily competitors could replicate your innovation.
Take, for example, a rideshare app. The concept of matching riders with drivers is broad and abstract—unlikely to meet patent eligibility on its own. But by layering in technical features, such as AI-driven routine prediction, route optimization for fuel efficiency, or matching based on environmental impact, you begin to build a framework of patentable elements around the core idea.
Checklist: Patent protection vs. trade secrets for AI inventions
- Is the invention patentable? Not just novel and non-obvious, but eligible under current guidelines.
- Can it be reverse-engineered? If yes, a patent might be better. If not, trade secret protection may be valuable.
- Are you seeking funding? Patents are tangible assets that signal innovation and attract investors.
- What’s the invention’s lifespan? If it’ll be obsolete in a few years, trade secrets may be more practical.
- Where do you want protection? Patent laws vary globally. What’s patentable in Europe may not meet patent eligibility requirements in the US.
- How important is confidentiality? Patents require public disclosure; trade secrets keep your innovation confidential.
Layered protection is often best
The most resilient IP strategies often layer protection, using patents for what’s enforceable and trade secrets for what’s best kept confidential. Copyrights and trademarks can also support broader innovation goals. As AI continues to evolve, so do the rules around protecting it. Staying agile, working closely with IP counsel, and aligning with a long-term AI intellectual property strategy are essential steps forward.
