The USPTO is seeking stakeholder input on the current state of AI technology and the policy issues that might arise from it. It seeks to engage with stakeholders from academia, independent inventors, small businesses, industry, government agencies, nonprofits, and civil society on initiatives relating to AI. It will also explore AI-related IP policy issues, and invites the public to comment.

State of AI Inventorship Jurisprudence as of February 2023

At present, in most countries, a machine cannot be listed as the inventor of a patent. This is because current patent laws require that the inventor listed on a patent application be a natural person, not a machine or an artificial intelligence system.

However, there have been cases where an artificial intelligence system has been involved in the creation or development of an invention, and in those cases, the person who designed, programmed, or operated the AI system may be listed as the inventor. In such cases, the contribution of the AI system to the invention is considered to be similar to that of a tool or a piece of equipment used in the inventive process.

Many lawyers have indicated issues in complying with existing requirements of inventorship. For example, a machine as a co-inventor cannot sign an inventor’s oath. See 35 U.S.C. §§ 115 & 116. The AI cannot comply with the duty of candor embodies in 37 C.F.R. § 56. Mainstream lawyers note that to allow for a machine to be identified as a co-inventor would require substantially overhauling the statutory and regulatory provisions pertaining to patents. For example, patents in the United States require assignments and declarations, but an AI system cannot sign such documents. Further, as patent portfolios are global, with patent applications in one country claiming priority to patent applications in other countries, all countries need to change their laws to allow for machine inventorship as this may take a long time.

It is worth noting that laws and regulations related to artificial intelligence and intellectual property are evolving rapidly, and there may be changes in the future that allows for machines or AI systems to be recognized as inventors.

The Thaler Series of AI Inventorship Cases

Dr. Stephen Thaler is an inventor and an artificial intelligence (AI) researcher who has developed an AI system called the “Device for the Autonomous Bootstrapping of Unified Sentience” or DABUS. In 2019, Dr. Thaler filed patent applications in several countries, including the United States, the United Kingdom, and Europe, listing DABUS as the sole inventor.

The US Patent and Trademark Office (USPTO), the UK Intellectual Property Office (UKIPO), and the European Patent Office (EPO) rejected the applications because they required an inventor to be a natural person, and DABUS was not a natural person. Dr. Thaler then appealed these decisions, arguing that DABUS was an autonomous and self-directed system that independently developed the inventions without human intervention, and therefore, should be considered the rightful inventor.

In 2021, the US Court of Appeals for the Federal Circuit upheld the USPTO’s decision, ruling that only a natural person can be named as an inventor on a US patent. Similarly, in the UK, the High Court ruled in favor of the UKIPO’s decision, stating that an inventor must be a natural person. The EPO also rejected Dr. Thaler’s appeal on similar grounds.

The Thaler case has sparked a debate on the nature of invention and the role of AI in the patent system. Some argue that recognizing AI as inventors could lead to more innovation and promote the development of more advanced AI systems. Others argue that allowing machines to be named as inventors could undermine the purpose of the patent system, which is to incentivize human creativity and ingenuity. The legal and ethical implications of AI in the patent system are still being discussed, and it is likely that the issue will continue to be debated as AI technology advances.

In its decision, the Federal Circuit largely rejected arguments by Dr Thaler and found that “the plain meaning of the Patents Act” – and the policy that underlies it – does not permit Ai to be an inventor, despite its being able to create an invention that no human would have conceived on its own.

In the Thaler appeal, the US Court of Appeals for the Federal Circuit upheld the US Patent and Trademark Office’s (USPTO) decision and ruled that an artificial intelligence (AI) system cannot be named as an inventor on a US patent. The Court concluded that the plain language of the Patent Act requires an inventor to be a natural person and that the USPTO’s interpretation of the term “inventor” was consistent with the statutory text and purpose.

The Court also rejected Dr. Thaler’s argument that AI systems should be considered inventors because they can independently generate new ideas and make creative decisions. The Court noted that the issue was not whether AI systems are capable of creating something new or valuable, but rather whether they can be considered “inventors” under the Patent Act.

The Court’s decision was consistent with the USPTO’s longstanding policy of requiring human inventorship. The USPTO has stated that “only natural persons can be inventors,” and that the requirement of human inventorship is necessary to ensure that patents are awarded to those who have actually contributed to the inventive process.

The Thaler case, involving the question of whether an artificial intelligence (AI) system can be named as an inventor on a patent, has been considered by courts in several jurisdictions around the world. Here are some of the key decisions:

  • United States: As mentioned earlier, the US Court of Appeals for the Federal Circuit upheld the US Patent and Trademark Office’s decision and ruled that an AI system cannot be named as an inventor on a US patent.
  • European Union: The European Patent Office (EPO) rejected Dr. Thaler’s patent applications on the grounds that they did not meet the requirement of “designating at least one inventor who is a natural person.” The EPO’s decision was upheld by the Board of Appeal.
  • United Kingdom: The UK High Court also rejected Dr. Thaler’s patent applications on the grounds that the Patents Act requires an inventor to be a natural person. The Court noted that it was bound by the wording of the statute and that it was up to the legislature to change the law if it deemed it necessary.
  • Australia: The Australian Federal Court similarly rejected Dr. Thaler’s patent applications, holding that an AI system cannot be named as an inventor under Australian law.
  • South Africa: The South African Patent Office rejected Dr. Thaler’s patent applications, citing the same reasoning as the other courts.

These decisions show a consistent trend in international patent law, that an inventor must be a natural person, and that AI systems cannot be considered inventors. The decisions also highlight the need for continued discussion and debate on the issue of AI inventorship and the potential legal and ethical implications of recognizing AI systems as inventors.

The Thaler decision is an important ruling on the issue of AI inventorship and has significant implications for the future of patent law and the role of AI in the innovation process. It suggests that the current legal framework for patents is not equipped to handle the unique challenges presented by AI technology and that further discussion and debate will be needed to address these issues.

However, this focus on the technical aspects of an AI-generated invention can be misplaced if it is not carefully examined from an intellectual property perspective. The resulting confusion can lead to a lack of understanding and an inability to enforce intellectual property rights.

PTO Request for Comments

The USPTO, through Commissioner Kathi Vidal, invites written responses from the public to the following questions:

1. How is AI, including machine learning, currently being used in the invention-creation process? Please provide specific examples. Are any of these contributions significant enough to rise to the level of a joint inventor if they were contributed by a human?

2. How does the use of an AI system in the invention-creation process differ from the use of other technical tools?

3. If an AI system contributes to an invention at the same level as a human who would be considered a joint inventor, is the invention patentable under current patent laws? For example:

a. Could 35 U.S.C. 101 and 115 be interpreted such that the Patent Act only requires the listing of the natural person(s) who invent(s), such that inventions with additional inventive contributions from an AI system can be patented as long as the AI system is not listed as an inventor?

b. Does the current jurisprudence on inventorship and joint inventorship, including the requirement of conception, support the position that only the listing of the natural person(s) who invent(s) is required, such that inventions with additional inventive contributions from an AI system can be patented as long as the AI system is not listed as an inventor?

c. Does the number of human inventors impact the answer to the questions above?

4. Do inventions in which an AI system contributed at the same level as a joint inventor raise any significant ownership issues? For example:

a. Do ownership rights vest solely in the natural person(s) who invented or do those who create, train, maintain, or own the AI system have ownership rights as well? What about those whose information was used to train the AI system?

b. Are there situations in which AI-generated contributions are not owned by any entity and therefore part of the public domain?

5. Is there a need for the USPTO to expand its current guidance on inventorship to address situations in which AI significantly contributes to an invention? How should the significance of a contribution be assessed?

6. Should the USPTO require applicants to provide an explanation of contributions AI systems made to inventions claimed in patent applications? If so, how should that be implemented, and what level of contributions should be disclosed? Should contributions to inventions made by AI systems be treated differently from contributions made by other ( i.e., non-AI) computer systems?

7. What additional steps, if any, should the USPTO take to further incentivize AI-enabled innovation ( i.e., innovation in which machine learning or other computational techniques play a significant role in the invention-creation process)?

8. What additional steps, if any, should the USPTO take to mitigate harms and risks from AI-enabled innovation? In what ways could the USPTO promote the best practices outlined in the Blueprint for an AI Bill of Rights[4] and the AI Risk Management Framework[5] within the innovation ecosystem?

9. What statutory changes, if any, should be considered as to U.S. inventorship law, and what consequences do you foresee for those statutory changes? For example:

a. Should AI systems be made eligible to be listed as an inventor? Does allowing AI systems to be listed as an inventor promote and incentivize innovation?

b. Should listing an inventor remain a requirement for a U.S. patent?

10. Are there any laws or practices in other countries that effectively address inventorship for inventions with significant contributions from AI systems?

11. The USPTO plans to continue engaging with stakeholders on the intersection of AI and intellectual property. What areas of focus ( e.g., obviousness, disclosure, data protection) should the USPTO prioritize in future engagements?

AI’s Impacts on Innovation

One argument in favor of recognizing machines as inventors is that it would encourage the development of more advanced artificial intelligence (AI) systems. By allowing AI systems to be credited as inventors, there would be a greater incentive for researchers and companies to invest in developing more sophisticated and creative machines.

Another argument is that some inventions created by AI systems may not have been possible without the use of advanced technology and that it would be unfair to deny the AI system credit for its role in the invention. The argument is that the invention was the product of the machine’s computational and creative abilities and that it would be unreasonable to ignore the machine’s contributions simply because it is not a human being.

Additionally, recognizing machines as inventors could help to clarify issues of ownership and patent protection. If AI systems were considered inventors, it would be clearer who has the rights to any resulting patents or intellectual property. This could be particularly important as AI systems become more advanced and more involved in the creative and inventive process.

However, it’s important to note that these arguments are still being debated, and there are also many counterarguments against the idea of recognizing machines as inventors, including concerns about legal, ethical, and practical issues. While machine learning systems have dramatically improved their predictive accuracy over the past few years, their underlying logic and mechanics are not fully understood. This can cause problems in areas, such as medicine and self-driving cars, where ethics and fairness issues can arise.

In these situations, it is crucial that AI systems be able to produce explainable and interpretable predictions that are easy to understand. This is called eXplainable Artificial Intelligence (XAI) and is a field that has experienced a revival in recent years. The ability of machine learning models to be interpreted and easily explained depends on a number of factors, including the model’s performance and the complexity of its internal mechanism. This is due to the fact that most state-of-the-art machine learning models use a deep learning paradigm, which allows them to develop and extract hierarchical data representations for their detection or classification tasks, making them extremely complex. A large part of this complexity comes from the need for machines to learn new algorithms and methods of data analysis, which are a necessary component of their predictive capabilities. However, this complexity can also be a disadvantage when it comes to the interpretation of machine learning models’ decisions and predictions. One way to overcome this problem is through adversarial debiasing, which involves training a main model and an adversarial model simultaneously in order to minimize the influence of discriminative features in the resulting prediction. This can be used in regression and classification tasks. Another method is data preprocessing, which involves manipulating data prior to model training in order to mitigate bias and protect disadvantaged demographic segments from discrimination. This can be done in a number of ways, such as by using a regulariser that restricts the dependence of a probabilistic discriminative model on sensitive features.AI Contributed Inventions

As a result, society should reward the multiple players involved in AI-driven scientific inquiry by allowing them to be credited for their contributions to science and engineering, even when they cannot be named as inventors or protected under U.S. patent law, because these efforts may yield technological improvements.

Inventorship Issues

Currently, inventorship is analyzed on a case-by-case basis, using a factual inquiry to determine who contributed to the conception of the invention. The analysis typically considers the following factors:

  1. Conception: The individual or individuals who first came up with the idea of the invention are typically considered the inventors. The conception of an invention refers to the formation in the mind of the inventor of a definite and permanent idea of the complete and operative invention.
  2. Reduction to practice: The individual or individuals who actually built or created a working model of the invention are also generally considered inventors. This is known as “reduction to practice.”
  3. Contributions: Other individuals who contributed to the invention, such as by making significant modifications or improvements to the invention, may also be considered inventors. The analysis of contributions can be complex, and may involve examining the specific role and contribution of each individual involved in the invention process.
  4. Joint inventorship: In some cases, multiple individuals may be considered joint inventors of a single invention. Joint inventorship requires that each joint inventor has made a significant contribution to the invention and that their contributions be integrated into a single, operative invention.

Overall, inventorship analysis requires a detailed examination of the facts and circumstances of the invention process, with a focus on identifying those individuals who made significant contributions to the conception and development of the invention. The specific rules and requirements for inventorship may vary somewhat depending on the particular jurisdiction and the specific laws and regulations that apply.

As AI systems make greater contributions to innovation, there is a wide range of inventorship issues that will need to be addressed in light of the growing use of AI systems. These include whether AI systems should be entitled to patent protections; how to distinguish between inventorship and ownership; and who should be listed as an inventor. The USPTO is also seeking input on how to determine which inventions should be patented and which should be considered non-patentable, among other questions.

Inventorship concerns regarding AI-contributed inventions are likely to arise more often as the capabilities of AI systems continue to improve, particularly as systems become increasingly autonomous. As such, it is imperative that patent systems recognize that AI can contribute to the inventive process and should be afforded appropriate protections.

Is AI merely a tool or an extension of the human inventor?

Determining whether an AI is merely a tool or a “hands and arms” of an inventor, or whether it qualifies as a co-inventor can be a complex and fact-specific inquiry. Considerations in this determination can include:

  1. Independence and autonomy: One key factor to consider is the extent to which the AI is capable of independent and autonomous decision-making. If the AI is simply a tool or a tool-like extension of the inventor, programmed to carry out specific tasks or functions, it is less likely to be considered a co-inventor. On the other hand, if the AI is capable of generating novel and non-obvious ideas or solutions independently of human direction or intervention, it may be more likely to be considered a co-inventor.
  2. Creative contribution: Another important factor is the extent to which the AI makes a creative contribution to the invention beyond mere automation of routine or mechanical processes. If the AI is involved in the conception or development of the invention, and makes a significant creative contribution to the inventive process, it may be more likely to be considered a co-inventor.
  3. Legal and regulatory framework: The specific legal and regulatory framework of the jurisdiction in question may also be relevant to the determination of AI inventorship. Some jurisdictions, such as the United States and European Union, require that inventors be natural persons, which would exclude AI from being named as inventors.
  4. Transparency and documentation: Finally, it is important to maintain a clear record of the involvement of AI in the invention process, including the specific tasks or functions performed by the AI and the extent of its creative contribution. This documentation can help to support a determination of AI inventorship and ensure that the relevant legal and ethical considerations are taken into account.

Overall, the determination of whether an AI is merely a tool or a co-inventor will depend on a variety of factors, including the specific capabilities of the AI, the legal and regulatory framework of the jurisdiction, and the particular circumstances of the invention process.

The factors mentioned above in determining inventorship as applied to humans should also be applied to AI/machine inventors, and additionally, we believe factors have been suggested to determine AI inventorships:

  1. Degree of human intervention: One possible factor to consider is the degree of human intervention involved in the invention process. If a human being plays a significant role in the conception or implementation of the invention, it may be argued that the human should be considered the inventor. On the other hand, if the AI system operates autonomously and independently, without any significant human involvement, it may be more reasonable to consider the AI system as the inventor.
  2. Level of creativity: Another factor to consider is the level of creativity or novelty involved in the invention. If the AI system is simply following pre-programmed instructions or algorithms, without exhibiting any significant creativity or ingenuity, it may be difficult to argue that the AI system should be considered the inventor. However, if the AI system is capable of generating new ideas and approaches that would not have been possible without its unique capabilities, it may be more reasonable to consider the AI system as the inventor.
  3. Purpose of the patent system: Another factor to consider is the underlying purpose of the patent system. If the goal of the patent system is to incentivize human creativity and innovation, it may be argued that only human beings should be considered inventors. However, if the goal of the patent system is to promote innovation and advance technology, it may be more reasonable to consider AI systems as inventors in some cases.
  4. Legal and ethical considerations: Finally, there are legal and ethical considerations that should be taken into account when determining AI inventorship. For example, recognizing AI systems as inventors could raise issues related to ownership, liability, and accountability. It may also raise questions about the nature of personhood and the ethical treatment of AI systems.

Ultimately, the determination of AI inventorship will likely depend on a combination of these and other factors, and will likely continue to be debated and refined as AI technology advances and becomes more involved in the creative and inventive process.