If you’ve ever needed a lawyer, your case may have gone through ChatGPT. Your name, your situation, the details you only told him. He didn’t tell you. And honestly, I understand why.
A lawyer I know well discovered ChatGPT for drafting. What used to take him three days now takes a few hours. He didn’t hesitate for a second: he pasted his cases in, client names included. Then one day he looked into it, found the rules of his profession, and stopped cold. His lawyer friends keep going.
So who’s right? Him, who stopped everything and went back to doing it all by hand? Or his friends, who save whole days every week?
My answer: neither. And to see why, you have to look at the part of this debate nobody looks at.
Three camps
AI is coming into law firms fast, hard, and everywhere. From what I see around me, lawyers fall into three camps:
- The ones who know, and don’t care. They put everything in, because it saves them a crazy amount of time.
- The ones who know, and stop. Like my lawyer.
- The ones who don’t even know they could use it, or how.
The first camp is much bigger than people think. And it doesn’t brag about it.
There’s a fourth camp. Almost empty right now. That’s the one that will win. I’ll come back to it.
What the rule really says
Quick side note, because a lot of lawyers get this wrong: AI is not banned. The French bar council put out its guide in March, and basically it says: use it if you want, but don’t put your clients’ info in a tool that can’t guarantee it stays secret. And consumer ChatGPT is exactly that: your data goes to the US, and depending on your settings, it can be used to train the model.
On paper, the rule makes sense. Professional secrecy is the core of the job. But in real life, it’s like asking an accountant who just discovered Excel to keep doing everything by hand. He’ll say yes. And he’ll open Excel the minute you turn your back.
Will anyone ever get in trouble for it? Honestly, I have no idea. What I know is what I see: the rule exists, and a lot of lawyers work around it.
The real problem isn’t moral
You could think the ones who work around the rule are just cheaters. I think it’s simpler than that.
Tools built for lawyers, hosted in Europe, already exist. But they cost more, you have to pick one, set it up, learn it. ChatGPT is already there, in your browser, for about twenty euros a month. Against a gain of three days, the clean option loses. Not because lawyers are dishonest. Because the clean option is harder.
It’s a product problem. And product problems get fixed.
The window is closing
Not tomorrow. I think it’ll take at least a year, maybe two. But it’s going to close.
The models are already amazing. For most of what a law firm needs, the output is close to perfect, and you won’t need much more. There will still be progress on very technical or very niche topics. But for the bulk of the work, the intelligence race is going to hit a ceiling nobody will notice anymore. European models like Mistral, hosted in Europe, aren’t there yet. They’ll need time to catch up. But they don’t need to be the best. They just need to reach the level that’s good enough. And then compliant tools will become as simple and as cheap as ChatGPT.
That day, the excuse disappears. And something happens that nobody sees coming.
The real losers
Logically, the people who worked around the rule should pay the price. It’s the opposite.
They took a risk, sure. But for a year or two, they learned. They know what to ask AI, what to fix, how to review. The day the compliant tool shows up, they change tools. Not jobs.
The real losers are the ones who did things right. My lawyer, who stopped everything to respect the rules, kept his hands clean. On his own, he would have lost a year or two of learning. The day the window closes, he would have started from zero, facing peers who got ahead while he waited. Same for the ones who didn’t even know: they discover everything on the same day, with nothing learned. Here, doing the right thing costs a lot.
And the fourth camp? It has the head start of the first group, without the risk. These are lawyers who learned to work with AI without ever putting client data in it. They anonymize their files. They build templates and review habits. They have the learning of the rule-benders, and the clean hands of my lawyer.
What we’re building
My lawyer, though, got lucky: he knows me. Together, we’re building a custom, compliant solution that gives him his days back without exposing his clients. And that fourth camp is exactly what we want to open to every lawyer at Hendy.
I won’t pretend to be neutral. Personally, I’m all-in: all my info goes through AI, and I sleep fine. But I’m not a lawyer: I’m not bound by professional secrecy. You are. So we’re building the best of both worlds: the power of the best models, without ever exposing client data. We’re not done, but it’s become a mission, and I’d rather a lot of us carry it. If this is on your mind, message me on LinkedIn. I’m happy to share what we’re learning.
What you do with this
So between my lawyer and his friends, who was right? Neither. He would have lost time on his own, they’re taking a risk. The one in the best spot is the one who learns without risking anything.
If you’re a lawyer, you have a year, maybe two. That’s plenty of time to build a method, as long as you start now. As for your clients’ data, it’s your call: take the risk knowingly, or learn to work without it.
If you’re a builder or an investor, look closely at what’s happening here. A whole profession is working around a rule because the clean version is still too expensive and too complicated. Whoever makes it as simple as ChatGPT has a whole market waiting.
And when the window closes, the people who learned while it was open will have nothing to catch up on.