Mauritz Kop Lecturer AI Regulation and Intellectual Property Law at CEIPI, University of Strasbourg
Strasbourg, France – We are pleased to feature insights from a lecture on "Intellectual Property and Ownership of AI Input and Output Data" delivered by Professor Mauritz Kop at the Centre for International Intellectual Property Studies (CEIPI), University of Strasbourg. This session was part of the University Diploma in Artificial Intelligence and Intellectual Property.
Rights and responsibilities pertaining to AI and data
Professor Kop, a Fellow at Stanford University and a strategic IP lawyer, shared his expertise on the rights and responsibilities pertaining to AI and data, offering both theoretical perspectives and practical tips at the current state of technological and legal development. The lecture aimed to equip attendees with a bird's-eye view of the intertwined key elements of this multidimensional topic.
Mauritz Kop Lecturer AI Regulation and Intellectual Property Law at CEIPI, University of Strasbourg
Main points of Professor Kop's lecture at CEIPI
Drawing directly from Professor Kop's lecture at CEIPI, the main points discussed included:
The Nature of Data Ownership (Anno 2022): A key takeaway highlighted that formal legal ownership of data, in the sense of property law, does not currently exist; rather, we operate with a concept of de facto or economic ownership. This legal uncertainty presents significant roadblocks.
Addressing Legal Roadblocks with Contracts: Current legal challenges and loopholes regarding the use of data in AI systems, particularly for machine learning training datasets, can often be navigated through strategically drafted data sharing contracts. These agreements can facilitate data sharing and limit liability in the absence of comprehensive horizontal legislation.
The Global Impact of EU Legislation: The EU Artificial Intelligence Act and the upcoming EU Data Act are poised to set global standards for AI and machine learning. It is anticipated that the norms and values embedded in this EU legislation will be exported to other democratic parts of the world, much like the GDPR framework did for privacy and data protection.
Strategic IP Management and AI Impact Assessments: Professionals can effectively map out intellectual property opportunities and risks by adopting an IP portfolio approach. This should be integrated as part of implementing an AI Impact Assessment (AIIA), a tool designed to ensure AI systems are legal, ethical, and technically robust through ex-ante and life-cycle auditing by multidisciplinary teams.
The Bigger Picture: Embedding Values in AI: The lecture emphasized the transformative power of AI and the critical importance of embedding democratic norms, standards, principles, and values into the design, architecture, and infrastructure of AI technologies from the outset. Technology is not neutral; as society shapes technology, technology shapes society.
IP Protection for AI Systems – A Portfolio Approach: A comprehensive strategy for protecting AI systems involves a "mixture of rights". Legally, an AI system comprises multiple components—such as the software/source code, training data, neural network, machine learning process, AI applications, hardware, and inference model—each potentially protectable by different IP rights (e.g., copyright, patents, trade secrets, database rights).
Challenges in Clearing AI Input Data: Significant hurdles exist in clearing AI input data, especially when training datasets consist of copyrighted works (music, photos, videos) or information protected by sui generis database rights (in the EU). Using uncleared data can lead to copyright infringement, while not using such data can slow AI development.
"Res Publicae ex Machina" – AI Output and the Public Domain: A novel concept introduced was "res publicae ex machina" (public property from the machine). This proposes that non-personal data autonomously generated by an AI system, without significant human creative contribution, should fall into the public domain, excluded from traditional IP protections.
The EU AI Act Deep Dive: The lecture provided an analysis of the EU Artificial Intelligence Act, highlighting its risk-based approach (pyramid of criticality), the objectives to ensure AI systems are safe and respect fundamental rights, and the stringent market entrance requirements for high-risk AI systems, - such as AI medical devices in healthcare and life sciences - including conformity assessments and CE marking.
Data Sharing Contracts in Practice: The structure and key clauses of data sharing agreements were discussed, emphasizing their role in protecting interests, avoiding misuse, and providing legal certainty. Important provisions include definitions, scope of use, data privacy, security, IP ownership, and warranties concerning IP clearance and regulatory compliance.
AI Impact Assessment (AIIA) Framework: The AIIA was presented as a practical checklist and code of conduct to guide the responsible and safe development and implementation of AI, in line with the European concept of Trustworthy AI (legal, ethical, technically robust).
Liability for AI Systems: The complex issue of liability for damages caused by smart robots and AI systems was addressed, noting the current reliance on existing national laws and legal doctrines in the absence of a specific AI liability framework, and contrasting EU and US approaches.
AI, data governance, and intellectual property law
Professor Kop's session underscored the dynamic interplay between AI advancement, data governance, and intellectual property law. It highlighted the necessity for legal professionals to be "double or triple educated" to navigate this complex field and for ongoing efforts to create legal frameworks that foster responsible innovation while addressing societal and ethical considerations.
The lecture concluded by stressing that AI literacy and awareness, continuous learning, and proactive legal strategies are essential for all stakeholders in the AI ecosystem. Thanks to Jean-Marc Deltorn for the kind invitation!