Artificial Intelligence (AI) is revolutionizing many industries, but its performance in the legal field has left much to be desired. Recent incidents have demonstrated that tools like ChatGPT, while useful for generating fast and coherent texts, cannot replace the experience and critical judgment of a professional lawyer. Why? Let’s explore the challenges and limitations that generative technologies face in legal contexts, as well as their potential applications with more moderate use.
Cases illustrating the failures of AI in law
In two recent episodes, one in the United States and one in the Benelux countries, the limitations of AI in the legal field were highlighted:
- The American case: A lawyer presented arguments based on nonexistent precedents generated by ChatGPT. These appeared legitimate, but turned out to be fabrications of the model. The result was disastrous: sanctions for the lawyer and a severe blow to his reputation.
- The “Artpenguin” case in the Benelux region: In a trademark dispute, the applicant relied on ChatGPT to defend its position against Penguin Books Limited. The AI-generated defense was irrelevant and confusing, even mentioning trademarks unrelated to the dispute. The result: the applicant lost the opposition, and the case served as an example of how not to use AI in legal contexts.
Why AI fails in legal contexts
Generative AI like ChatGPT is a powerful statistical system that produces plausible texts based on linguistic patterns, but its design has crucial limitations for use in law:
- Lack of precision and veracity: AI doesn’t “know” what’s true; it simply generates content that sounds plausible. This leads to dangerous fabrications, such as nonexistent laws or irrelevant precedents.
- Inability to understand context: Laws and legal arguments require a deep understanding of the rules, precedents, and unique circumstances of each case. AI lacks this understanding.
- Lack of ethical judgment: A lawyer must consider not only technical aspects but also moral and strategic ones. AI, by its nature, lacks an ethical compass.
- Strategy errors: Presenting a case requires anticipating the other party’s moves, something AI cannot do effectively.
When can AI be useful?
Despite its limitations, AI can be a valuable tool if used correctly. Some practical applications include:
- Initial drafts of documents: AI can help structure arguments or prepare drafts that then need to be reviewed by an expert.
- Simplifying legal concepts: Explaining complex terms in accessible language.
- Basic research: Extract general information on applicable laws and regulations.
However, even in these cases, the result should be verified by a lawyer to avoid critical errors.
The lawyer as an irreplaceable strategist
A good lawyer is not only a legal expert, but also a strategist who understands the bigger picture. Like a chess master, a lawyer assesses the opposing party’s possible responses, anticipates risks, and adapts their approach based on the context. AI, on the other hand, is more comparable to a chess manual: useful as a reference, but unable to play the game itself.
For example, in the “Artpenguin” case, an experienced attorney would have considered the risks of confusion from the perspective of the average consumer and presented arguments based on solid precedent and psychological analysis, something ChatGPT simply cannot do.
Risks of relying exclusively on AI in law
Blindly relying on AI tools in the legal field can lead to serious problems:
- False expectations: The apparent precision of generated texts can mislead both lawyers and clients.
- Waste of resources: A poorly prepared document can lead to court rejections and additional costs.
- Loss of trust: Serious mistakes can damage the reputation of a professional or company.
we believe that AI can be a great ally in the legal field if used with caution and as a complement to human expertise. While it can save time and offer new insights, it should never replace a lawyer’s critical judgment. As the BOIP stated in its assessment: “Promptness can be improved.” We also believe the best way to leverage AI is to use it as a tool, not a substitute
