The AI Endgame in IP
Judge Plager wrote, in a 2018 Federal Circuit Dissent1):
“Today we are called upon to decide the fate of some inventor’s efforts, whether for good or ill, on the basis of criteria that provide no insight into whether the invention is good or ill. Given the current state of the law regarding what inventions are patent eligible, and in light of our governing precedents, I concur . . . even though the state of the law is such as to give little confidence that the outcome is necessarily correct.”
In another decision2), The Federal Circuit unanimously concluded:
“We agree . . . that the district court clearly abused its discretion in evaluating [defendant’s] transfer motion and denying transfer.”
Justice Scalia, in a dissent3), famously declared:
“It follows, as night follows the day, that only valid patents can be infringed. To talk of infringing an invalid patent is to talk nonsense.”
These examples hint at the challenges we face as humans in establishing clear and consistent rules for creating and enforcing patented intellectual property. When judges themselves express frustrations with the state of the law, it’s worth considering whether there might be a better way.
I am a big fan of British science fiction author, Neal Asher. In his books, Asher imagines a far-flung future wherein virtually all of humanity is ruled by artificial intelligences. Being non-human, the AIs of Asher’s novels lack avarice, prejudice, and cowardice. They are imperfect, but far more perfect than humans, and far more capable of ruling. Humanity flourishes under the AIs’ rule (but of course there are dissenting factions–and therein lies some great plot lines).
Patent judges, attorneys, and examiners, being human, are undeniably fallible. Modern patent law and modern technology have both become incredibly complex creating expensive litigations that are ripe for abuse by bad-faith actors. What if we could apply AI technology to this system to bring down these costs and at the same time improve fairness and consistency?
AI in Patent Drafting and Examination
Let’s assume that AI will soon be capable of performing searches of prior art more thoroughly than any team of humans. Given that ability, an AI can draft a patent application—based on a detailed invention description—that would be grantable upon filing. If that AI is appropriately certified by the USPTO, there would be no need for a traditional examination because the AI itself would ensure that the claims distinguish the invention from the global corpus of prior art, that the written description is complete and enabling, and that the claims comply with § 101 jurisprudence. This significantly reduces the cost of drafting and virtually eliminates the cost of patent examination.
Because filing and examination are suddenly so cheap, we can expect an exponentially greater number of registered patents. But, importantly, a much higher percentage of these will be valid. For example, a new technology is created that might form the foundation for a whole new product line. Instead of patenting the most important features of the invention, the innovator can every possible aspect that is novel.
AI in Litigation and Enforcement
AI can also greatly reduce the cost of litigation and enforcement. Questions of validity are more readily addressed and resolved using AI, thereby reducing the scope of patent disputes primarily to questions of infringement and damages. However, although the scope of litigation may be narrowed, the volume is potentially much higher due to the much greater volume of inexpensive registered patents.
To manage this surge, we might see the rise of intermediary licensing AIs (ILAs). These ILAs would represent companies, with full access to all documents, email records, engineering files, and sales data. A patent owner seeking a license from a target company could simply submit its patent to that company’s ILA, which would generate a fair license offer or declare the patent invalid or uninfringed. If the patent owner disagreed, litigation would remain an option, but the credibility of the ILA will carry weight.
If two competitors face off, each could submit its patents to the other’s ILA. The ILAs might then negotiate a cross-license. If one party wants an injunction rather than a license, the ILA would determine whether the asserted patent truly reads on the competitor’s product, and if it does, and the competitor lacks leverage of its own IP, the ILA could advise the competitor to agree to halting the sale, and additionally recommend product changes to avoid the asserted patent.
The Human Factor
Obviously, this is not a panacea. A lot of attorneys, examiners, paralegals, etc., and possibly even judges, will lose their jobs. The post AI economy is a serious economic problem—far broader than the our small bubble of intellectual property law in which we live—that society writ large will need to address. But for the attorneys that remain, there will still be a need for professionals who understand both the technology and the complexities of patent law.
While AI systems will significantly reduce certain responsibilities—such as prior-art searches, drafting discovery requests, and reviewing large volumes of discovery material—they will also create new demands. Who will certify these systems for reliability and fairness? Who will ensure that AI-generated patent claims align with a company’s strategic goals and ethical commitments? And who will stand ready to challenge or defend AI-driven decisions in court? These emerging needs require legal experts who possess both deep subject-matter expertise and the human judgment needed to oversee and guide AI. In this environment, patent law will shift from manual labor and routine tasks to higher-level scrutiny, strategy, and advocacy. Even in the universe imagined by Neal Asher, humans remain essential deciders: the spaceships are sentient AIs, but they still have human captains.