AI Patent Protection and Litigation: Key Takeaways for Innovators and Companies

Artificial intelligence is transforming industries and redefining how innovation is created, deployed, and protected. As companies accelerate development of generative models and machine-learning systems, questions around patent eligibility, inventorship, and enforcement have taken center stage. Regulators, courts, and the US Patent and Trademark Office must now determine how traditional patent frameworks apply to technologies that evolve autonomously, process massive datasets, and, in some cases, generate their own outputs.

For patent stakeholders, understanding the contours of patent protection for artificial intelligence (AI) technology is critical. In a recent Technology Marathon webinar, our team looked at how recent developments reveal both expanding opportunities for protection and new litigation complexities surrounding infringement detection, invalidity defenses, and discovery of proprietary data and models.

AI INVENTIONS REMAIN PATENTABLE BUT MUST MEET CLEAR LEGAL THRESHOLDS

• AI-based inventions can qualify as patentable subject matter if they meet the requirements of 35 USC sections 101, 102, 103, and 112
• Claims must do more than describe abstract algorithms; rather, they must demonstrate a technical solution to a technical problem, such as a specific model design, training process, or interface improvement
• The US Patent and Trademark Office (USPTO), under current leadership, has shown growing support for well-drafted AI claims that show concrete computer improvements or practical applications

PATENT STRATEGY SHOULD REFLECT GLOBAL DIFFERENCES

• Europe, China, and Japan generally recognize AI inventions as patentable when tied to technical implementations, but there are subtle differences:
• For example, European examiners assess whether the invention contributes to technical character, while China emphasizes a technical solution using technical means that yields a technical effect
• Across jurisdictions, showing that an AI system solves a technical problem is key to success

AI CANNOT BE AN INVENTOR, BUT HUMAN INVOLVEMENT MUST BE DOCUMENTED

• Under Thaler v. Vidal and USPTO guidance, only natural persons qualify as inventors
• February 2024 guidance from the USPTO reaffirmed that AI-assisted inventions are patentable only with significant human involvement (i.e., inventors must meaningfully contribute to each claim, and simply operating or owning an AI system is not sufficient)
• Applicants should document who designed, trained, or prompted the system that produced the inventive concept

EXPECT MORE LITIGATION AS PATENT GRANTS INCREASE

• Rising allowance rates for software and AI technologies are likely to drive more infringement suits
• Proving or disproving infringement can be difficult when source code and training data are confidential or evolving

DETECTING AND PLEADING INFRINGEMENT REQUIRES FLEXIBILITY

• Because many AI processes are hidden, plaintiffs often rely on public materials, such as marketing language, technical papers, or documentation referencing neural-network layers or training methods
• Courts have accepted such evidence at the pleading stage, recognizing plaintiffs may not yet access the accused code
• Companies promoting AI capabilities should assume public statements may later appear in a claim chart

DRAFT CLAIMS WITH LITIGATION IN MIND

• Robust specifications and well-defined claim language strengthen patents against indefiniteness and enablement challenges
• Broader terminology can capture competitor activity but must remain grounded in sufficient disclosure
• Focusing on server-side functions, where infringement by competitors is more likely, can be strategically advantageous

PREPARE FOR SECTION 112 AND DISCOVERY CHALLENGES

• AI patents frequently face indefiniteness and enablement scrutiny
• Discovery in AI litigation is data-intensive, often requiring preservation of evolving datasets, source-code versions, and algorithmic parameters
• Early litigation holds, proportionality arguments, and protective orders are key to managing cost and confidentiality

LOOKING AHEAD

Patent law’s core principles remain steady, but AI technology is changing how those principles are applied. Applicants must clearly document human inventorship, disclose technical detail sufficient for enablement, and show tangible technical improvements, while litigants must adapt infringement and discovery strategies to fit these dynamic and oftentimes opaque technologies.

As the USPTO and courts refine their approaches, companies integrating thoughtful patent strategy into every stage of AI development will be best positioned to protect innovation and mitigate litigation risk.