Does patent protection create value in AI-based companies?

Hours

Thanks to an invitation from the University of Silesia, which is implementing the B-Kinetic project, Stanisław Rogoziński had the opportunity to co-host a meeting with Dr. Tomasz Konopczyński (hetalox.com) entitled: “Does patent protection create value in AI-based companies?”

Many startups operating in the AI space are wondering whether patenting their solutions makes sense. On the one hand, it is pointed out that protection of IT solutions is limited and varies by jurisdiction, a significant part of the solution must be disclosed in a patent, and obtaining and maintaining a patent is costly. On the other hand, obtaining a patent signals novelty, originality, and applicability of a solution. This increases credibility in discussions with investors and clients and can also be used for marketing purposes. It facilitates conversations with potential technology buyers. It remains debatable whether patent protection in the case of AI solutions creates real barriers to entry.

Tomasz Konopczyński spoke clearly in favor of patenting AI solutions. He cited, among others, the example of HeartFlow, which established cooperation with Siemens thanks to its patents, as well as the Polish company Hemolens Diagnostics. He emphasized the defensive role of patents, which can be used to block larger competitors or protect against lawsuits. In addition, the discussion covered the following topics:

  1. Protection strategies – what to disclose in the patent claims, what to describe in general terms, and what to keep as a trade secret. We also discussed how researchers can reconcile scientific publications with patent protection. Some aspects of an invention are better described using value ranges, others through exemplary values, while some should not be disclosed at all.
  2. What “AI patent protection” actually means – this may include elements such as model architecture, parameters and hyperparameters, or the dataset used to train the model.
  3. How patent protection creates company value – it was emphasized that everything depends on the startup’s business strategy. If the goal is to maintain freedom to operate or to sell the solution, patents can be very useful. However, some companies prefer an open-source model and make their code publicly available. Hugging Face and LangChain have operated as open source from day one. None of these companies patent their solutions, but for example Hugging Face was sued in a patent lawsuit in January 2025 for alleged patent infringement.
  4. Additional topics included the level of costs associated with patenting, regional differences, and issues related to patenting inventions created by AI.

In summary, the more a solution meets the criteria of an invention or the more specific its application, the more justified patenting becomes. If we decide to patent our solution, it is crucial to carefully plan the budget for filing the application, extending it to selected regions, and then maintaining the protection. Poorly designed protection can be more harmful than having no patent at all. In the case of AI solutions, it is worth dividing our discovery into elements that we disclose in a patent or publication and those that we keep confidential.