BRAZIL
August 21, 2025
The Brazilian Patent and Trademark Office (BRPTO) has officially opened a public consultation on the draft guidelines for the examination of patent applications related to artificial intelligence (AI).
The invite has been published on the Official Bulletin No. 2850, dated August 19th, 2025.
The consultation period is 60 days from publication – October 18, 2025.
This initiative marks an important step in shaping the future of AI patent protection in Brazil, allowing stakeholders, industry, academia, and practitioners to contribute to the framework that will guide the examination of AI-related inventions.
The first draft prepared by the Brazilian PTO has the following key points:
1 - Subject matter not considered as inventions (Art. 10 IP Law):
- The mere use of AI as a tool to arrive at an invention does not make the subject matter patentable. Hence, an invention must: (i) solve a technical problem, (ii) provide a technical solution, and (iii) produce a technical effect. Databases, raw datasets, or training data are excluded from patent protection, as they are considered information presentation or subject to copyright protection.
- Computer programs, per se, are not inventions, but computer-implemented inventions (CII) with a concrete technical effect may be patentable.
2 - Sufficiency of disclosure (Arts. 24 and 25 of IP Law):
- The specification must be clear and complete to allow reproduction by a skilled person without undue experimentation.
- Challenges such as the “black box” nature of AI models do not prevent compliance if the invention is described with enough technical detail.
- Claims must explicitly indicate the technical application of AI (e.g., “Method for facial recognition using a neural network...”), not merely the AI model itself.
3 - Inventive step:
- The person skilled in the art is defined as someone with knowledge in both AI techniques and the technical field of application.
4 - Situations generally considered obvious:
- Mere automation of known processes with AI.
- Simple combinations of known AI techniques without unexpected effects.
- Routine parameter tuning (e.g., number of layers, learning rate) unless producing a surprising technical effect.
- Substituting one known AI model for another with a similar function, unless adapted in a non-obvious way.
5 - Potentially inventive situations:
- AI is working in synergy with specific hardware implementations, leading to unexpected technical effects.
- Novel data processing techniques (collection, normalization, vectorization, training environment setup) that are not routine and yield technical effects.
The draft and further details are available on the INPI Portal and on the Participa + Brasil platform. (https://www.gov.br/inpi/pt-br/servicos/patentes/consultas-publicas/Minutadasdiretrizes.pdf)
Contributions must be submitted exclusively through the official electronic form: https://forms.gle/8AKLcnv5dgCtQEod7
If you have any questions or require additional information, please do not hesitate to contact us.