Artificial Intelligence Under the European Patent and Copyright Law

Artificial Intelligence (AI) now pervades inventive and creative workflows, raising questions for European patent and copyright law. This thesis adopts an in-depth doctrinal and policy perspective across the European Patent Convention (EPC) and the EU copyright acquis. Its objectives are fourfold: (1) to translate the technology of AI into concepts that matter for legal analysis; (2) to assess whether and how AI systems can themselves be protected; (3) to evaluate how law should treat AI-influenced inventive and creative outputs; and (4) to propose reforms that safeguard incentives while preserving access and transparency. Chapter 2 provides a concise technical primer on AI – core methods (machine learning, neural networks, deep learning), data-dependence, adaptivity, and the “black-box” effect – so later chapters build on realistic capabilities rather than hype. Chapter 3 examines European patent law. It explains when AI systems qualify as patentable inventions – typically based on the methodology of assessment of computer-implemented inventions – and how mixed claims are assessed for novelty and inventive step by focusing on features that contribute to a technical effect. It also addresses plausibility and sufficiency, emphasizing the value of disclosing algorithms, data flows and validation tests to support reproducibility. The chapter then evaluates AI-assisted inventive output and the DABUS debate, concluding that, at present, AI functions as a tool within human-directed problem-solving and that inventorship remains with human contributors who devise the inventive concept and recognize the relevance of results. Chapter 4 turns to EU copyright. It maps protection for AI systems via software and database copyright and, where appropriate, sui generis database rights. It then evaluates AI-influenced outputs against the EU originality standard (“author’s own intellectual creation”) and the need for human free and creative choices to be perceptible in the expression across the creative process (conception, execution, redaction). Where sufficient human control is absent, outputs fall outside copyright as works, although related-rights frameworks may sometimes apply. Chapter 5 sets out a measured legal agenda. For patents: maintain the technical-character gateway; require claims to articulate the AI element’s technical purpose and effect; appraise inventive step with a “would-not-could” lens; and support sufficiency with transparent model-validation evidence. For copyright: reaffirm the human-authorship threshold; clarify that purely AI-generated outputs are unprotected as works. The thesis positions today’s AI as a powerful but non-autonomous enabler. Protecting genuinely technical AI innovations and clarifying attribution for AI-assisted results is achievable within existing EPC doctrine, while EU copyright should continue to ground protection in human creative choice. A cautious, evidence-led path – resisting premature recognition of machine inventors or authors– best preserves Europe’s innovation incentives.