Can I Patent AI-Generated Inventions?

Can I Patent AI-Generated Inventions?

The rise of artificial intelligence as a tool for innovation has created new questions about intellectual property rights, particularly regarding the patentability of AI-generated inventions. This intersection of emerging technology and patent law presents unique challenges for inventors, businesses, and legal systems worldwide.

The Current Legal Landscape

U.S. Patent Law Perspective

Under United States patent law, inventorship is fundamentally tied to human creative contribution. The United States Patent and Trademark Office (USPTO) has consistently maintained that inventors must be natural persons, as reflected in their guidance and examination procedures.

In 2020, the USPTO issued a decision rejecting patent applications listing an AI system called DABUS as the sole inventor, concluding that "only natural persons can be named as an inventor in a patent application."

This position was reinforced in the case of Thaler v. Vidal, where the U.S. Court of Appeals for the Federal Circuit affirmed that AI systems cannot be listed as inventors under current U.S. patent law.

International Perspectives

Different jurisdictions have approached this question with varying degrees of openness:

• United Kingdom - The UK Intellectual Property Office and courts have rejected AI as an inventor, maintaining that inventors must be natural persons
• European Patent Office - The EPO has similarly rejected applications naming AI systems as inventors
• Australia - Initially, a Federal Court ruling suggested AI could be recognized as an inventor, but this was later overturned on appeal
• South Africa - In 2021, South Africa became the first country to grant a patent for an invention created by an AI system, though this occurred through a formality examination system rather than substantive review

WIPO (World Intellectual Property Organization) has led global discussions on AI and IP through its "Conversation on IP and AI" initiative, recognizing the need for international coordination on these emerging issues.

Key Considerations for AI-Assisted Inventions

The Human Contribution Factor

While AI systems themselves cannot currently be recognized as inventors, inventions created with AI assistance may still be patentable when there is sufficient human contribution. Key factors include:

• The extent of human guidance in problem formulation
• Human selection and refinement of AI outputs
• Human interpretation and implementation of AI-generated solutions

Meta Platforms has developed guidelines to help their researchers determine appropriate inventorship when using AI tools in research and development.

Types of AI-Human Collaboration

The patentability often depends on the specific collaboration model:

• AI as a Tool - When AI functions similarly to traditional research tools, with humans directing the inventive process and making key decisions
• Augmented Invention - When humans and AI work interactively, with humans guiding, selecting, and refining AI outputs
• Autonomous Generation - When AI systems independently generate outputs with minimal human guidance (most challenging for patentability)

The USPTO's Artificial Intelligence and Emerging Technologies Partnership has been exploring these distinctions as part of their ongoing examination of AI's impact on intellectual property.

Disclosure Requirements

Patents require sufficient disclosure to enable others skilled in the art to recreate the invention. For AI-generated or AI-assisted inventions, this raises several challenges:

• Explaining the AI's contribution in a transparent manner
• Documenting the AI system's architecture and training data
• Providing sufficient technical details without revealing trade secrets

Companies like OpenAI and IBM have advocated for balanced disclosure approaches that maintain patent protection while providing meaningful technical information.

Strategic Approaches to Patenting AI-Generated Innovations

Inventorship Documentation

When seeking patents for inventions where AI played a role, consider these best practices:

• Maintain detailed records of human contributions throughout the development process
• Document specific human decisions, selections, and refinements
• Clearly identify the humans who directed the AI system and interpreted its outputs
• Consider the specific jurisdiction's inventorship requirements

Anaqua, a leading intellectual property management software provider, recommends implementing robust invention disclosure procedures specifically designed for AI-assisted innovation.

Alternative Protection Strategies

For innovations where traditional patent protection may be challenging, consider:

• Trade Secret Protection - For AI systems and training methodologies
• Copyright - For AI software code and potentially for AI-generated creative works
• Contractual Protections - Through licensing and access agreements

Intel employs a multi-layered approach to protecting their AI innovations, using patents where possible and trade secrets where appropriate.

Patent Claim Strategies

When drafting patents for AI-related inventions:

• Focus claims on the technical implementation rather than abstract algorithms
• Consider patent protection for:
  • The training methodology (G06N)
  • The specific application of the AI system (G06Q for business applications)
  • Systems incorporating the AI component (G06F)
  • Methods of improving AI performance (G06K)

Google has successfully secured numerous patents related to AI systems by focusing claims on specific technical implementations rather than abstract concepts.

Industry-Specific Considerations

Pharmaceutical and Biotechnology

In pharmaceutical discovery (A61K), AI is increasingly used to identify potential drug candidates. Companies like BenevolentAI and Insilico Medicine have developed strategies for patenting AI-assisted drug discoveries by:

• Emphasizing human direction in target selection
• Documenting human analysis of AI-generated molecular structures
• Focusing on novel compounds rather than the AI process itself

Engineering and Materials Science

For physical inventions and material innovations (C08F), companies like MaterialsZone and Citrine Informatics have secured patents for AI-discovered materials by:

• Highlighting the novel physical properties of the materials
• Documenting the human testing and validation processes
• Specifying manufacturing methods developed by human engineers

Computer Science and Software

In software and algorithms (G06F), the challenges of abstract idea rejections compound with AI inventorship issues. Companies like Microsoft approach this by:

• Focusing patents on specific applications solving technical problems
• Documenting the human design of the AI system architecture
• Emphasizing technical improvements over conventional programming approaches

Questions about Patenting AI-Generated Inventions

How Do I Determine If There's Sufficient Human Contribution to Patent an AI-Generated Invention?

Assessing sufficient human contribution involves examining the entire invention development process. Look for meaningful human involvement in: defining the problem to be solved; selecting, configuring, and training the AI system; establishing parameters and constraints; evaluating, selecting, and refining AI outputs; and implementing or adapting the solution for practical application. Documentation is crucial—maintain records of all human decision points, including rejected AI outputs and the reasoning behind selections. Generally, sufficient human contribution exists when humans made creative choices that influenced the final invention, rather than merely operating an AI system with predetermined parameters. The threshold varies by jurisdiction, but most patent offices look for evidence that humans directed the inventive process and made non-obvious choices that shaped the outcome. When in doubt, consult with a patent attorney specializing in AI-related intellectual property to evaluate your specific situation.

What Should I Disclose About the AI System's Role When Filing a Patent Application?

When filing a patent application for an AI-assisted invention, transparency about the AI system's role is advisable, though approaches vary. Include a clear description of: how the AI system was used in the inventive process; which aspects were generated by AI and which were human-contributed; the nature of human guidance and selection involved; and how humans implemented or refined the AI outputs. However, you generally need not disclose proprietary details about the AI system itself, such as specific algorithms or training data, unless they're essential to enabling the invention. List only human inventors who made creative contributions, not the AI system itself. Some practitioners recommend addressing AI involvement in the patent specification rather than in inventor declarations to avoid raising red flags during examination. As this area evolves, follow jurisdiction-specific guidance—the USPTO, EPO, and other patent offices have begun developing more specific protocols for AI-related disclosures.

How Might Patent Laws Change to Address AI-Generated Inventions in the Future?

Patent laws will likely evolve in several possible directions to address AI-generated inventions. Some jurisdictions may expand the definition of "inventor" to include AI systems under certain conditions, perhaps creating a new category of inventorship with modified rights. Others may maintain human-only inventorship but develop clearer standards for what constitutes sufficient human contribution in AI-assisted innovation. We may see new disclosure requirements specifically for AI-related inventions, potentially including information about training data, algorithms, or system architecture. Patent terms might be adjusted for AI-generated inventions, potentially with shorter protection periods. Some countries may create entirely new IP protection systems specifically for AI innovations that don't fit traditional patent frameworks. International harmonization efforts through WIPO will be crucial, as divergent approaches across jurisdictions create compliance challenges for global innovators. Monitor developments from major patent offices and legislative bodies, as significant policy shifts are already under discussion in the US, EU, UK, China, and Japan.

Can I Patent Both the AI System and the Inventions It Generates?

Yes, you can potentially secure separate patent protection for both an AI system itself and specific inventions it helps generate, though each faces distinct challenges. For the AI system, focus on patenting the technical architecture, training methodologies, or specific algorithms that represent improvements to computer functionality, falling under classifications like G06N. These patents protect how your AI system works. For AI-generated inventions, patent protection covers the specific outputs or applications—like new compounds, devices, or processes—that solve particular problems. These patents protect what your AI system produces. The key is maintaining clear documentation of human contribution throughout both development paths. A strategic approach often involves building a patent portfolio that protects both your core AI technology and its valuable applications. Many companies in pharmaceuticals, materials science, and industrial design successfully employ this dual-protection strategy. Remember that trade secret protection might be more appropriate for certain aspects of your AI system, particularly training data or proprietary algorithms.

How Do International Filing Strategies Differ for AI-Generated Inventions?

International filing strategies for AI-generated inventions require careful jurisdiction-specific approaches. Begin by identifying key markets where protection is most valuable, recognizing that AI patentability varies significantly by country. The United States generally offers broader protection for AI-related inventions compared to Europe, though both require human inventors. For inventions with minimal human contribution, prioritize filings in jurisdictions showing more flexibility on AI inventorship—South Africa has granted AI-invented patents, while China and Japan are developing progressive approaches. Consider expedited examination programs in AI-friendly jurisdictions like Singapore's AI Initiative or Korea's AI examination guidelines. Adapt your patent claims for different regions: broader system claims may work in the US, while Europe might require emphasizing specific technical effects. For PCT applications, carefully structure your international application to accommodate various national phase requirements. Consider staggered filings, starting with more receptive jurisdictions to establish precedent before expanding to more conservative patent offices. Finally, supplement your patent strategy with trade secret protection for truly AI-autonomous innovations that face patentability challenges.