Algorithms are mathematical instructions used to process data and solve problems. These innovations can be protected by patents, such as Google’s PageRank algorithm, which is central to the company’s search engine. Alternatively, some organizations opt to protect their algorithms through trade secrets, like Coca-Cola’s AI-driven flavor algorithms, which are not publicly disclosed. A key challenge in protecting AI-related inventions, such as algorithms, is determining whether they are eligible for protection under existing IP frameworks. Since algorithms are often based on mathematical concepts, they may not be considered inventive or novel enough to be granted patent protection.
To make AI algorithms more eligible for patent protection, it is essential to claim them in a manner that goes beyond mathematical concepts and highlights their practical application or technical solution provided . For example, instead of focusing solely on the mathematical aspects of a machine learning algorithm, emphasize how it improves the efficiency of a specific process, such as optimizing energy consumption in a smart grid system.
Datasets are collections of information that AI systems use to learn and improve. They can be protected through copyrights, like the ImageNet database, which has significantly advanced the field of image recognition. On the other hand, some companies may choose to protect their datasets as trade secrets, such as proprietary financial datasets that are valuable assets for investment firms.
Strategies for making datasets more eligible for patent protection can include demonstrating the innovative and non-obvious methods employed in collecting, curating, or organizing the dataset . For instance, a patent application for a dataset could describe a unique approach to gathering real-time traffic data from a network of sensors and aggregating it in a manner that significantly improves traffic predictions.
3. Trained Models:
Trained models refer to AI systems that have been refined using specific algorithms and datasets. These models can be protected by patents, such as IBM’s Watson, a powerful AI system for natural language processing. Alternatively, some companies may protect their trained models as trade secrets, such as Tesla’s autopilot software.
To enhance the eligibility of trained AI models for patent protection, it is crucial to claim them as part of a broader inventive system or method. An example of this approach could involve a patent application for an AI-powered diagnostic tool that incorporates the trained model, emphasizing how the model interacts with various components of the system, such as sensors and user interfaces, to deliver accurate and rapid diagnoses.
4. AI-created Works:
AI-generated music, art, or literature may be eligible for copyright protection if they involve creative modifications and arrangements made by humans, as indicated by the U.S. Copyright Office. In general, to be eligible for copyright protection, the work must be original, creative, and fixed in a tangible medium.
5. AI-generated Designs:
Designs created by AI, such as fashion designs, product packaging, or user interfaces, may be protected under design rights if they are new, unique, and have an individual character, which grant exclusive rights to the creator of the design. However, the key challenge is proving that the AI-generated design meets these requirements, considering that the design is a product of machine learning algorithms rather than human input.
6. AI-related Trademarks:
Trademarks can be used to protect AI-generated marketing slogans, logos, or other brand identifiers. This can help organizations maintain a unique brand image and prevent unauthorized use of their AI-generated intellectual property. One concern in this area is to ensure that the AI-generated marks are unique and do not inadvertently infringe upon others’ IP rights.