Artificial Intelligence 101: Key Challenges and Opportunities in Patent Law and Generative AI

Artificial intelligence (AI) has transformed industries globally, and its influence on intellectual property (IP) is no exception. While predictive AI systems have been used for decades, generative Artificial Intelligence 101 represents a newer development with far-reaching implications, especially in patent law. Generative AI systems, such as ChatGPT, DALL-E, and LLaMA, leverage machine learning models to analyze patterns from existing data and produce new, creative content.

This blog explores four pivotal aspects of generative AI in patenting: patent inventorship, AI-generated prior art, and statutory hurdles.

Patent Inventorship and Artificial Intelligence 101: Who Owns the Innovation?

Generative Artificial Intelligence 101 introduces a fascinating dilemma regarding patent inventorship. The Federal Circuit mandates that only human beings can qualify as inventors. However, determining whether a human’s contribution, aided by generative AI, meets the inventorship threshold often presents ambiguity.

For example, an inventor might rely on generative AI to solve complex problems or create innovations. If the AI system significantly influences the outcome, determining whether the human or the AI system “conceived” the invention can lead to disputes.

The United States Patent and Trademark Office (USPTO) has issued guidance to address such challenges, including the following principles:

• Using an AI system doesn’t preclude a human from being an inventor.
• Recognizing a problem or formulating a plan alone doesn’t constitute conception.
• Significant contributions to inventorship must go beyond routine experimentation or practice.
• Ownership of an AI system doesn’t automatically make its owner the inventor of AI-generated outputs.

However, key questions remain:

• What qualifies as a “significant contribution” when AI systems assist?
• How do we differentiate between an AI tool’s and a human inventor’s roles?

Answering these questions requires further legal clarity and potentially new regulations to address the nuanced interplay between humans and generative AI.

AI-Generated Prior Art: Friend or Foe?

The sheer ability of generative AI to produce volumes of technical content introduces challenges in prior art analysis. Prior art refers to evidence that an invention is already known and not patentable.

Generative AI’s capacity to create complex outputs poses two risks:

1. Overwhelming Volumes: Generative AI could flood the system with vast amounts of prior art, increasing the time and cost for inventors and legal teams to evaluate its relevance.
2. Technical Inaccuracy: AI-generated prior art might lack the necessary technical depth or practical utility, complicating its applicability in patent decisions.

Proposed Guardrails for AI-Generated Prior Art

To address these issues, courts, and policymakers might consider additional requirements:

• Enablement Test: AI-generated references should not automatically qualify as prior art unless they enable a person skilled in the art to replicate invention.
• Human Oversight: AI-generated references should undergo human review to ensure they meet conception standards.

Implementing these safeguards could minimize the risk of irrelevant or technically flawed prior art. This, in turn, incentivizes inventors to pursue their patent applications confidently.

Eligibility Under 35 U.S.C. § 101: Can AI Innovations Be Patented?

The patent eligibility of AI-driven inventions often faces scrutiny under 35 U.S.C. § 101, which defines what constitutes patentable subject matter. Since generative AI technologies involve algorithms and computational processes, they may fall under abstract ideas, which are typically non-patentable.

However, Artificial Intelligence 101 innovations can still be eligible for patents if they demonstrate:

• A practical application of the abstract idea.
• Significant improvements to computer functionality or a specific technological field.

For instance, Patent Trial and Appeal Board (PTAB) reversed a § 101 rejection for a machine learning patent by highlighting its advancements in memory usage and classifier accuracy. This example underscores the importance of how inventions are framed in patent applications.

To improve patent eligibility for AI-driven inventions, innovators should focus on:

• Demonstrating unique data preparation techniques.
• Highlighting technological advancements in Artificial Intelligence 101 models.
• Presenting compelling use cases that show practical benefits beyond abstract ideas.

Statutory and Regulatory Hurdles

Generative AI also introduces regulatory challenges in patent law. These include:

1. Signing Submissions: USPTO regulations require submissions to be signed by natural persons. Generative AI tools capable of automated submissions could inadvertently violate this requirement.
2. Confidentiality Issues: Using online generative AI tools to input patent-eligible subject matter could trigger public disclosure rules, impacting patent validity.
3. Verification Standards: Legal professionals using AI tools must ensure the accuracy of AI-generated results, as they are accountable for the integrity of their submissions under USPTO standards.

To navigate these challenges, inventors and patent practitioners should:

• Avoid relying solely on generative AI tools to draft or submit applications.
• Maintain strict confidentiality when using AI systems for sensitive IP-related work.
• Double-check AI-generated content for compliance with USPTO requirements.

Conclusion

Generative Artificial Intelligence 101 is undeniably a transformative force, but its integration into patent law raises complex challenges. The legal landscape is evolving rapidly from questions of inventorship to the validity of AI-generated prior art. As courts and regulatory bodies refine their approaches, inventors and practitioners must stay informed and adaptable.