AI and Contracts in Finland: Confidence Is Not Competence

The use of generative Artificial Intelligence tools such as ChatGPT and Microsoft Copilot are becoming increasingly popular in legal and non-legal worlds alike. They can produce polished contract drafts in seconds. AI contracts can and often do look exactly like human-made contracts – and that is the problem. Thus, a question arises: what is the standing of AI contracts in the context of the Finnish legal system?  

The Dunning-Kruger Effect and AI Contracts  

Psychologists Dunning and Krueger have shown that people with limited knowledge of a field tend to overestimate the quality of their own work because they lack the expertise to notice what is missing. The same trap applies to anyone without legal training reviewing an AI-generated contract. The document reads well, and the risks hiding it do not announce themselves. Only a real person with legal training can recognize the many potential pitfalls in contracts.

LLMs Know Law — Just Not Finnish Law 

Large Language models (LLMs), such as Chat GPT-4, are predominantly trained in English-language text from common law jurisdictions, above all the United States and the United Kingdom. Thus, Anglo-American contract law is deeply embedded in how these models think and write.  

Finnish contract law, however, comes from the Nordic civil law tradition. Finnish contract law acts such as The Contracts Act, the Sale of Goods Act, the Employment Contracts Act and their interpretive case law operate on different default rules and assumptions than anything an AI has meaningfully trained on. When you ask an AI to draft a contract “under Finnish law”, you typically receive a common law reasoning with a Finnish governing law clause bolted on top.  

A concrete example of this is that Finnish courts interpret ambiguous contract terms against the party who drafted them. An AI that does not understand this principle will draft clauses that create the exact liability that the client wanted to avoid. Our firm has written about this rule in detail in in Epäselvän sopimusehdon tulkinta laatijansa vahingoksi (in Finnish). 

Practical Implications 

As such, limitation of liability clauses built on US assumptions may not hold up in a Finnish court; a non-compete drafted in the American style may conflict with the Employment Contract acts, for example. These concerns are not theoretical in nature — they surface when a dispute arises.  

Moreover, AI tools are not designed to distinguish generally reputable sources from unreliable ones, nor do they grasp the doctrine of sources of law across different jurisdictions. Unreliable sources inevitably lead to incorrect conclusions. This limitation is not expected to disappear any time soon.

We Use These Tools — and Still Cannot Trust Them 

Using AI in your work can be extremely helpful in sketching structure and completing routine tasks more efficiently. However, it needs to be kept on a short leash. AI does not detect the legal issues that matter, nor does it understand the context or the risks of the deal or the relationship of the parties involved.    

Our legal experts assist entrepreneurs and small and medium-sized businesses with contract drafting and review. Book a free initial call with us here or call +358 1031 92030.

Antti Piispanen
attorney, partner