Menyhárd Attila, professor at Eötvös Loránd University (ELTE) and head of its Artificial Intelligence and Technology Law postgraduate program, says artificial intelligence has become a fundamental legal concern, not only a technological, business or educational topic. The arrival of AI forces a rethink of basic legal concepts such as liability, discrimination, legal personhood, data processing, product liability and corporate compliance.
Why ELTE started the program and how it was built
According to Menyhárd, the idea for the program originated years ago, partly driven by legal and technical questions around self-driving cars. The biggest challenge was creating a curriculum practically from scratch: there was no preexisting model to copy. To launch a university-level program, ELTE needed transferable, structured knowledge, and that foundation had to be developed for AI law.
Lawyers and engineers must learn each other’s language
A central point of the conversation was that lawyers and computer scientists must learn to communicate. Menyhárd noted this is not entirely new for lawyers — after the regime change they had to work with accountants, later with economists; now they must engage with engineers and IT specialists. Developers need to know the legal frameworks they'll operate within, while lawyers need to understand what the developers intend to build. Without that shared understanding, an iterative and potentially harmful cycle can emerge, especially because a development may reach the market only two to three years after it begins while legal regulation can change in the meantime.
Could AI become a legal person?
One of the most discussed topics was whether AI could ever be a legal person. Menyhárd argues that legally this is not impossible — the law is flexible and can create legal subjects from various entities — but the real question is whether society wants it. Such a change would be a social, moral and political decision. Currently the law divides the world into legal subjects and legal objects: either someone is an owner, or something belongs to someone. If AI became a legal person, questions would arise about who retains ownership, who would lose property rights, and whether developers would be willing to build technologies they could not sell. Menyhárd therefore says it is premature to rush into legally personifying AI unless there is a clear societal demand.
What generative AI can and cannot do for legal work
The discussion also addressed the limits of generative AI. Menyhárd stressed that large language models can produce convincing, well-written text, but that does not mean they "understand" it. In legal practice this distinction matters: generative AI can help with simple tasks such as basic property-protection questions or document review, but it cannot independently construct a legal strategy for a specific case, perform nuanced risk analysis, or develop fact-specific solutions.
Legal education must teach adaptability, not just tools
Menyhárd emphasized that legal education must adapt: it is impossible to know exactly which technologies a law student enrolling today will use in five years. Therefore education should focus less on a particular tool and more on preparing students for change, developing autonomous professional judgment and fostering the ability to collaborate across disciplines.
Knowledge of AI law is no longer optional
The professor was clear: knowledge of AI law is no longer optional. Artificial intelligence is already integrated into corporate operations, the economy and social processes. Those who want to comply with legal and regulatory requirements cannot ignore this field.
Details from the programme timestamps
- How did ELTE’s Artificial Intelligence and Technology Law postgraduate programme start? (00:45)
- Why must lawyers and computer scientists learn to communicate? (02:59)
- Who applies to the AI-law postgraduate programme? (10:54)
- Can artificial intelligence be a legal person? (15:57)
- How does AI change the work of lawyers? (30:11)
- How should future law students be prepared? (36:13)
- Why is knowledge of AI law no longer optional? (39:49)
The full conversation is available in the original episode.
