The Law of Artificial Intelligence
Legal, ethical, and societal implications of AI developments; EU AI Act; designing AI systems compliant with the law.
The two seminars require preparation and active small-group work. Seminar groups and instructions are published on Canvas. Worth attending: the seminar discussions show up on the exam and you cannot really study them from the slides.
Learning objectives
At the end of the course, the students will: understand the legal, ethical, and societal implications of AI developments. be able to write informed opinions about AI legislation. be trained in acquiring a set of complex legal AI-related topics in a restricted period of time. be able to work together as a team to identify not only the technical but also the legal issues with the design of AI systems. be familiar with basic knowledge about the law that applies to AI systems, including the law specifically focusing on AI systems such as the EU AI Act. be able to identify legal risks when designing AI systems. be in a position to design AI systems compliant with the law. be able to use legal rules to protect their AI systems. be able to reflect critically on the impact of legal rules on technology. be able to express their own motivated opinions and contribute to the writing of better legal rules.
The course is organised around five lectures and two practical seminars, packed into roughly three weeks at the end of the year. It starts with Lessig's four modalities of regulation and the relationship between law and code, then builds the EU AI Act from its risk-based rules (including Articles 5–15) through the separate capabilities-based rules for general-purpose AI models (Articles 51–55). The final lectures examine exemptions, regulatory sandboxes, competition and innovation, and the wider legal stack around AI: the GDPR, DSA, DMA, copyright, product liability and related EU rules. The seminars turn that theory into practice by having students redraft Article 5 prohibitions and work on computational compliance for copyright text-and-data-mining opt-outs.
How it is taught. Lectures are discussion-led rather than slide-reading, using a Socratic style, so the professor expects you to have done the mandatory readings and to actually talk. The two seminars put you in small teacher-assigned teams to work on real drafting problems.
The material is rewritten every year. The professor updates the readings, the examples and the emphasis annually to track what is happening in AI regulation. That is what keeps the course engaging, and it is also why old summaries and question banks from previous cohorts are unreliable here.
Assessment. A single on-campus multiple-choice exam worth 100% of the grade, covering the mandatory readings and preparation material as well as the lectures and seminars. A mock exam is released partway through, and there is one resit. Students who take the resit are not eligible to graduate with honours.
Worth knowing before you sign up. This is one of the lighter courses in the programme in terms of workload, and the course manual states plainly that if you attend the sessions, come prepared and participate, you will be fine for the exam. No programming, no maths, and no group deliverable to coordinate.
Assessment
A multiple-choice exam on 24 June accounts for 100% of the final grade. The exam covers the mandatory readings and preparation materials as well as the lectures and seminars. A mock exam is available for practice. There is one resit opportunity.
Teaching methods
Five discussion-led lectures and two practical seminars. Students prepare the assigned material before class and work in small, teacher-assigned teams during the seminars.
Literature
Mandatory readings, podcasts, legal texts and preparation materials are provided through Canvas.