What AI Will Never Conquer in IP Analysis

I spend a lot of my time thinking about what AI can do. That's sort of the whole deal when you're building an AI platform. You wake up, you read about what the models can do now, you think about what they'll be able to do next quarter, and you try to stay ahead of it. It's exhilarating in the way that standing in the middle of a highway might be exhilarating.

But lately I've been thinking more about the inverse. Not what AI can do, but what it can't. And specifically, what it will never be able to do in the world of intellectual property. I don't mean "never" in the way people said we'd never have self-driving cars. I mean "never" in the way that the law, and the nature of human creativity, and the basic architecture of how ideas become protected innovations actually work. There are walls here that aren't technical. They're structural. And no amount of compute is going to knock them down.

The Law Says So (And It Means It)

Let's start with the most literal answer. Under U.S. patent law, only a natural person can be an inventor. This isn't some dusty, neglected corner of the statute. It was tested directly in Thaler v. Vidal, where the Federal Circuit ruled in 2022 that AI systems cannot be listed as inventors on patent applications. Stephen Thaler, a computer scientist, had filed two patent applications naming his AI system, DABUS, as the sole inventor. The court said no. The Patent Act requires a human being.

Source: Holland & Knight, "The Final Word? Supreme Court Refuses to Hear Case on AI Authorship and Inventorship," March 2026. https://www.hklaw.com/en/insights/publications/2026/03/the-final-word-supreme-court-refuses-to-hear-case-on-ai-authorship

Thaler pushed the same argument into copyright law, and in March 2026 the Supreme Court declined to hear his appeal in Thaler v. Perlmutter. That leaves the D.C. Circuit's ruling intact: human authorship is required for copyright protection. The district court in that case called human authorship a "bedrock requirement." The courts aren't being coy about this.

Source: Mayer Brown, "Supreme Court Denies Cert in AI Authorship Case," March 2026. https://www.mayerbrown.com/en/insights/publications/2026/03/supreme-court-denies-review-in-ai-authorship-case

And then in November 2025, the USPTO issued revised inventorship guidance that doubled down on this principle. The new guidance treats AI the same way you'd treat a laboratory flask or a calculator. It's a tool. It can help, but it cannot conceive. Conception, in patent law, is defined as forming a "definite and permanent idea" of the complete invention in the mind of the inventor. That word, "mind," is doing a lot of heavy lifting. It means a human mind.

Source: Holland & Knight, "The Human Element: USPTO Clarifies Inventorship for AI-Assisted Inventions," February 2026. https://www.hklaw.com/en/insights/publications/2026/02/the-human-element-uspto-clarifies-inventorship

So right off the bat, the legal system has drawn a line. AI cannot invent. It can assist, it can process, it can identify patterns across millions of documents. But the spark of conception that the law requires? That has to come from a person.

Strategy Is Not a Search Problem

Here's where it gets more interesting, and where I think a lot of people in this space get confused. The hardest parts of IP work were never about data retrieval. Yes, prior art searches used to take months. Yes, sifting through patent databases was painful and tedious and expensive. AI has changed that (and for what it's worth, that's exactly what we built Paseo to do). Paseo can analyze an invention against over 200 million patents and journal articles and return a full report in about ten minutes. That used to be weeks of billable hours.

But here's the thing. Getting that information faster doesn't answer the harder question: what do you do with it?

Michael Porter wrote in Competitive Strategy about how companies need to match their strengths and weaknesses to the structure of their industry. He talked about finding positions where competitive forces are weakest, about devising strategies that don't just cope with those forces but actually alter them. That kind of thinking applies directly to IP strategy. Knowing what patents exist in a space is only the beginning. Deciding which claims to pursue, which white spaces to target, which competitors to watch, which markets justify the cost of filing internationally, and when to keep something as a trade secret instead of patenting it at all: those are judgment calls. They require context, experience, risk tolerance, and an understanding of business goals that goes well beyond what any model can provide.

Source: Michael E. Porter, Competitive Strategy: Techniques for Analyzing Industries and Competitors, Free Press, 1980. (Uploaded project reference.)

A patent portfolio is not a collection of documents. It's a competitive weapon. And wielding it well requires the same kind of strategic imagination that Porter described. AI can tell you where the white space is. It can't tell you whether that white space is worth occupying.

Privilege Requires a Pulse

This one is newer, and it's a big deal. In February 2026, a federal judge in the Southern District of New York ruled in United States v. Heppner that documents a defendant created using a consumer AI platform were not protected by attorney-client privilege. The defendant, facing federal fraud charges, had used a publicly available AI tool to draft legal documents on his own before sharing them with his attorneys. The court applied the standard three-part privilege test and found the documents failed on every element. The AI platform is not an attorney. There was no reasonable expectation of confidentiality (the platform's terms of service explicitly allowed data retention and disclosure to third parties, including government authorities). And the purpose wasn't to obtain legal advice from a lawyer.

Source: Perkins Coie, "Federal Court Rules Client's Use of Generative AI Is Not Privileged," February 2026. https://perkinscoie.com/insights/update/federal-court-rules-clients-use-generative-ai-not-privileged

The court emphasized that all recognized legal privileges presuppose a trusting human relationship with a licensed professional bound by fiduciary duties. AI meets none of those criteria. It has no license to lose, no ethical code it's bound to, no duty of loyalty. As the Harvard Journal of Law & Technology put it in a recent essay, AI systems are "neither persons nor professionals: no licensure, oath, or sanction applies."

Source: Harvard Journal of Law & Technology, "Against an AI Privilege," November 2025. https://jolt.law.harvard.edu/digest/against-an-ai-privilege

This matters enormously for IP work, where sensitive invention disclosures, business strategies, and competitive intelligence flow between inventors and their patent counsel every day. The relationship between an inventor and their attorney isn't just a formality. It's the protective structure that makes honest, thorough IP development possible. You can't replicate that with a chatbot, no matter how good the chatbot is.

Someone Has to Be Responsible

Edward O. Wilson, in his book Consilience, wrote about the obstacles to creating truly human-level artificial intelligence. He identified what he called "the functional obstacle" and "the evolutionary obstacle." The first was about complexity: rational thought doesn't emerge from computation alone but from continuous exchanges between body and brain, shaped by emotional controls that regulate attention and goal-setting. The second was about history: human cognition is the product of millions of years of evolution in environments that shaped who we are.

Source: Edward O. Wilson, Consilience: The Unity of Knowledge, Vintage Books, 1998. (Uploaded project reference.)

Wilson was writing about AI in general, but his point translates perfectly to the legal profession. Patent attorneys don't just apply rules to facts. They exercise professional judgment shaped by training, experience, ethical codes, and accountability to bar associations, courts, and clients. When a patent attorney drafts claims, they're making dozens of judgment calls about scope, language, prosecution history, and enforceability. If they get it wrong, there are consequences. They can be sanctioned, sued for malpractice, or disbarred. That accountability structure is what makes the whole system trustworthy.

AI has none of that. A language model that generates draft patent claims (and there are tools doing this right now) bears no responsibility if those claims are too narrow, too broad, or fail to capture the actual invention. The attorney who relies on that output is still the one holding the bag. The ABA's Formal Opinion 512 makes this clear: attorneys remain accountable for all work product, regardless of how it was created.

Source: American Bar Association, "AI and Attorney-Client Privilege: A Brave New World for Lawyers," ABA Business Law Section, 2024. https://www.americanbar.org/groups/business_law/resources/business-law-today/2024-september/ai-attorney-client-privilege/

So Where Does AI Actually Fit?

I want to be honest about something, because I think it matters. I'm the founder of an AI platform. I have every incentive to oversell what AI can do. But I also think the best case for tools like Paseo is made by being clear about what they're for and what they're not.

What Paseo does is collapse the information-gathering phase. It runs your idea through novelty analysis, obviousness scoring, competitive landscaping, prior art identification, and white space mapping, all against a database of over 200 million patents and journal articles. It does this in about ten minutes. That's real, and it's transformative for the speed of innovation cycles.

But the report that comes out the other end is not a strategy. It's intelligence. It's the map, not the decision about which direction to march. Paseo itself recommends that a lawyer review any draft claims it generates before proceeding. That's not a disclaimer for legal cover. It's the reality of how this work needs to be done.

Klaus Schwab, writing about the Fourth Industrial Revolution, asked whether AI's growing role in decision-making could lead to a world where humans begin to act like robots. It's a fair question. But in IP, I'd argue the opposite is happening. The faster AI can deliver the data, the more clearly the human elements stand out. The judgment, the strategy, the relationships, and the accountability. Those aren't being automated away. They're being revealed as the core of what makes IP work actually work.

Source: Klaus Schwab, The Fourth Industrial Revolution, World Economic Forum, 2016. (Uploaded project reference.)

AI is going to keep getting better at processing information. It's going to keep getting faster. And it's going to keep making it easier for inventors and IP professionals to understand the landscape around their ideas. But conception is still a human act. Strategy is still a human decision. Privilege still requires a human relationship. And accountability still demands a human being on the other end.

The law isn't confused about this. We shouldn’t be either.