Hello everyone,
today I'd like to share some thoughts on the use of AI in the legal profession.
In recent weeks, well-known Bitcoiners have been actively making predictions about which professions will become obsolete next and how quickly AI will replace them.
This is a big topic. Nobody can avoid it.
So, our old boy Jack Mallers
@jack mallers
has listed several professions, including mine, namely lawyers. He has justified his claims. His main arguement: only the artists, the creative (at least this is my reading of this words), will survive.
I have already expressed my opinion on this matter, which can be read elsewhere.
Today, I'd like to comment on a foreword (to a new issue of a journal) by the editor-in-chief of one of the major German law journals (NJW, 03.09.2025 - Tobias Freudenberg). The foreword mentions some phenomena in the legal profession in Germany that I find amusing.
I quote:
"The dangers of uncontrolled use of AI are even more evident in a recent decision of the Higher Regional Court of Celle (29.4.2025 – 5 U 1/25, BeckRS 2025, 12085). The court notes that the defendant has cited several incorrect sources, including one from NJW-RR. We must know: in fact, the cited page contains a decision of the Federal Court of Justice, not the Higher Regional Court of Munich. The cited case number does not exist, at least it cannot be found in beck-online or juris. The same applies to other decisions cited by the defendant.
The relevant judgment of the Federal Court of Justice (NJW 2011, 3575) was not cited by the defendant. At another point, the court criticizes the defendant for an argument that is "in diametrical contradiction to their own previous statements". Although the court does not explicitly assume that AI was used, it is likely that this was the case.
Those who think that the plaintiff will be triumphant after this 'brainwashing' of the defendant are mistaken. "According to the current state of affairs, the defendant's appeal will be successful," the court states. For the plaintiff, things get even worse: although their claim would have been valid based on the defendant's statements, the plaintiff's own lawyer did not explicitly adopt these statements, even as a subsidiary claim. Based on their own statements, the claim is unfounded.
Here, too, it is pure speculation whether and to what extent AI may have contributed to this dilemma. However, there are indications that this may be the case. Fabricated sources are not the only risk associated with the use of AI, as Tobias Voßberg has pointed out in the aforementioned editorial: "Whether an argument follows three marginal numbers later that undermines one's own litigation strategy, whether the source is tactically useful or rather harmful, cannot be decided by AI in the future either. This responsibility remains with us."
All of this corresponds to our experiences over the past 10 months or so. We have come into contact with texts that, in our opinion, were likely generated with the help of AI.
On the one hand, this comes from clients who are very proactive and think they can help us, sometimes even instruct us. And they send us clever-sounding texts.
On the other hand, it comes from colleagues who represent the opposing party.
In both cases, the effort required from us is the same. Not only do we have to scan the text for consistency, but we are also increasingly forced to verify the sources, All INDICATED SOURCES. Which is new to the extent.
And we completely share the opinion cited above!
Caution is the mother of the porcelain box! As a German proverb puts it.
Don't trust any other writer, especially not when AI is evident in the text.
My impression (the authoer of the foreword uses the phrase "Blind trust is not justified, among other things, because AI hallucinates"):
AI wants to impress, and I have expressed this opinion several times before. At the expense of correctness. It's astonishing how smoothly a derivation or argumentation can be read. Then I take a deep breath and start checking... And often, at the end, there's just hot air.
Well, AI optimists and maximalists may say: "Yes, that's today. But it will all get better. These mistakes will soon no longer occur."
Okay, that may be true. But I'm not talking about next year. I'm talking about today. Because I work today. And today, I can make (expensive) mistakes if I offer poor service to my clients.
In short: see above - be aware of the dangers in using AI - e.g. the porcelain box
funny times though
