Comparative Patent Laws for AI related inventions

What is Patentable when it comes to AI

US:

1. § 101 defines the patent eligibility for AI related applications. The U.S. courts use the Alice/Mayo test for abstract ideas:

- Is the claim directed to an abstract idea (e.g., a mathematical method)?

- If so, does it add “something more” that transforms it into a patent-eligible application (a practical,

technical application)?

2. AI-assisted inventions must still have significant human contributions for patentability.

Europe:

1. Under Article 52(2)-(3) EPC, mathematical methods and computer programs “as such” are excluded from patentability.

2. Exception: If an AI-related invention has a technical character, it may be patentable. The EPO requires that the “mathematical method” (e.g., an algorithm) contributes to a technical solution of a technical problem.

3. The “technical effect” doctrine is central: the invention as claimed must result in a technical benefit (for example, improved image processing in a medical device, enhanced control in a robot, or better signal processing).

4. AI / ML models are treated as a subset of “computer-implemented inventions.”

5. Key Case Laws:

-T 1173/97 (“Computer program product/IBM”): A seminal decision in which the EPO held that a

computer program product is patentable if, when run on a computer, it produces a further technical

effect beyond simple software-hardware interaction.

-T 641/00 (“COMVIK / Two identities”): This decision provided a framework for assessing inventive step

when there are both technical and non-technical features. Only the features contributing to the technical

character are considered for inventive step.

-T 258/03 (“Auction method/Hitachi”): Holds that a method involving technical means can be patentable,

but one must carefully assess the claim for technical contribution.

India:

1. Section 3(k) of the Indian Patents Act 1970 explicitly excludes: “a mathematical or business method or a computer programme per se or algorithms” from being considered inventions.

2. Exception: If the computer program (or algorithm) produces a technical effect or “technical contribution,” the invention may still be patentable.

3. AI-related inventions may be patentable if they are tied to hardware or provide a concrete technical improvement and are not claimed as a mere algorithm or computer program.

4. According to the Section 3(k) of the Indian Patents Act 1970, AI inventions which fall under Algorithms, Mathematical models and computer programs per se are excluded unless the invention shows a technical effect or technical contribution, such as:

   (a) Improved hardware functioning

   (b) Reduced processing time

   (c) Enhanced system security

(d) A technical solution to a technical problem

Note: AI-assisted invention is judged solely on established patentability criteria (novelty, inventive step, industrial applicability, sufficiency), not on whether the AI used public, private, or mixed data.

Inventorship & Disclosure/Enablement

US:

1. Inventors must be human. AI systems cannot be named as inventors.

2. Requires a “person” who contributes conceptually.

3. The patent applicant needs to show that they were in possession of the invention.

4. Applicants need to provide detailed information on training data, parameters, model architecture, or even sample code to satisfy the disclosure requirements.

5. The USPTO applies a “significant contribution” test (based on a longstanding legal test from Pannu v. Iolab Corp.) to determine whether a human’s input is enough to qualify for inventorship.

6. The Pannu test considers:

   (a) the nature and significance of the human’s contribution,

   (b) how it relates to the invention as a whole, and

   (c) whether the person did more than just explain known concepts.

7. Simply “overseeing” or “dominating” an AI system isn’t enough. For instance, if a human just provides a problem statement or supervises, that may not satisfy significant contribution.

8. The USPTO does not require a new, formal duty to disclose that AI was used in creating the invention. However, if an examiner has reason to question inventorship (e.g., whether the named humans actually made enough of the inventive contribution), some information about how AI was used (like inputs / outputs) may be material and could be submitted.

9. USPTO allows AI-assisted innovation but not “automatic” patents for inventions purely generated by machines.

10. As per the patent law, the disclosed patent application should enable a “person skilled in the art” to make and use it.

Europe:

1. Inventors must be human. AI systems cannot be named as inventors.

2. European patent application does not need to show a special “human contribution” for patentability.

3. The origin of the inventive idea (whether from AI or human) is irrelevant to the assessment of patentability.

4. A human can be named as inventor if they made a material contribution to the conception of the invention, even when using AI tools.

5. The sufficiency of disclosure requirement (Article 83 EPC) applies. The application must enable a skilled person to carry out the invention.

6. The EPO often expects detailed technical disclosures (architecture, training, parameters, data) so that one skilled in the art can reproduce the technical solution.

7. Unlike US, in Europe, the Person Skilled in the Art (PSA) is treated as a team of experts rather than one individual, meaning the expected level of technical knowledge is higher. So, the disclosed patent application should enable them to make and use it.

India:

1. Indian law presumes a person (natural or legal) as inventor and there is no legal recognition of AI systems as inventors.

2. Inventors must be human. AI systems cannot be named as inventors.

3. A human must be named even if AI contributed technically.

4. Applicants in India often need to draft their claims carefully to emphasize how the AI interacts with hardware, or how it solves a “technical problem” in a system.

5. India does not have a formal “significant human contribution test” like the USPTO, but the practical effect is similar: someone must take responsibility for the inventive concept.

6. Patentability is judged on the invention itself, not on how it was conceived.

7. The Indian Patent Office does not examine:

   (a) Who (human or AI) generated ideas

   (b) Whether an AI contributed scientifically

   (c) The degree of human involvement in creating the output

8. Applicants do not need to disclose that AI was used or contributed for the invention.

9. So, India aligns with Europe on Inventorship and not the US.

10. In India, enablement is judged from the viewpoint of a single Person Skilled in the Art (PSA), and the applicant must also disclose the best method known for performing the invention.

    Note: Enablement requires disclosure of how to make and use the invention and how the inventor came up with it. That means, the inventors don't need to teach others how to exactly replicate the creative process, and rather only the technical product or method. So, AI is just a tool, and inventors need to disclose only what is necessary to enable a skilled person to practice the invention, not how the AI tool works.

    Summary

US: Inventors must be human and AI systems cannot be named as inventors. AI-assisted technical inventions (not just abstract ideas) must still have significant human contributions for patentability. There is no requirement that applicants need to disclose how much of the invention was made by AI or by a human during patent examination. However, if an examiner has reason to question inventorship (e.g., whether the named humans actually made enough of the inventive contribution), some information about how AI was used (like inputs / outputs) may be material and could be submitted.

Europe: Inventors must be human and AI systems cannot be named as inventors. However, European patent application do not need to show a special “human contribution” for patentability. So, a named human inventor must exist, but their personal contribution is not part of the patentability examination. There is no requirement that applicants need to disclose how much of the invention was made by AI or by a human during patent examination.

India: Applicants need to draft their claims carefully to emphasize how the AI interacts with hardware, or how it solves a “technical problem” in a system. Patentability is judged on the invention itself, not on how it was conceived. There is no requirement that applicants need to disclose how much of the invention was made by AI or by a human during patent examination.

Questions to think of:

1. How do major patent jurisdictions define “inventor” in their patent laws, and how is this applied in practice when AI contributes to or autonomously produces inventions?

2. What are the substantive patentability criteria applied to AI-related inventions in different jurisdictions (e.g., technical character, inventive step, enablement)?

3. How do patent offices assess disclosure (specification) requirements for AI inventions, especially regarding model architecture, training data, human interaction and their algorithmic complexity?

4. What legal and policy reforms could better align patent systems with the reality of AI-driven innovation?