Take all unreasonable measures

We can't dismiss extreme possibilities anymore, in either the geopolitical or the legal worlds. We need to recognize radical change and prepare new strategies for exceptional times.

Monday morning, I spoke with a legal journalist about the application of Generative AI to lawyer formation. I began by noting, as I usually do, the difficulty of making concrete observations about Gen AI right now, given that it’s evolving so quickly — nobody really knows its current or future capabilities in the legal space.

“There’s a wide spectrum of potential outcomes here,” I said. “At the near end of the spectrum, Gen AI plateaus tomorrow, and we’ll only have to deal with what it can already do. At the far end, Gen AI becomes so powerful that it basically replaces most lawyer work. Those are the extreme outliers, of course,” I continued. “The likely outcomes are somewhere between those endpoints.”

That’s true, as far as it goes. But there’s a reasoning error nested in there to which I’ve fallen victim. We name the two extreme endpoints as our parameters — but then we ignore them and focus on what’s in the middle. The extremes are also possibilities. If you get angry at the meteorologist because she forecast a 10% chance of rain and it ending up pouring on your picnic, she’ll remind you (pedantically but correctly) that she was right: Rain was possible, and it rained. 10% is not 0%.

Similarly, I’ve discounted the “AI replaces most lawyers” possibility, and there’s a couple of reasons for that. One is simply that it still feels outlandish to me: Such an extreme scenario ought to be many years away. But the other, if I’m being honest, is that treating the idea seriously would cost me credibility with the conservative and skeptical audience (you good folks) that I’m trying to reach. Not only do lawyers have great difficulty imagining a world without lawyers, but they also find the suggestion kind of offensive. They’d tune me out if I talked about it, and disregard much else that I had to say.

Humans have serious mental blocks when I comes to their own annihilation. People who predict the end of the world are still caricatured as walking around in sandwich boards because we don’t want them to be right. Our minds recoil from the prospect of our own demise, so we push back hard against anyone who makes us think about it.

As we’ve all discovered over these last two tumultuous weeks, however, ignoring extreme possibilities because they’re remote and make us uncomfortable is a dangerous business.

Let's stick with AI for the moment. Hours before my Monday morning conversation, OpenAI released Deep Research, a Gen AI system that can conduct research with the depth and nuance of human experts at machine speed. Combining the capabilities of AI Reasoners and AI Agents, Deep Research has entered the charts at what Ethan Mollick calls the performance level of a novice PhD student. It goes into exhaustive detail. It provides links to real websites for its sources. It’s basically a researcher in a box, and while it’s still limited and imperfect, it’s a major advance in Generative AI.

How well would it function in the legal space? Colin Lachance and Bob Ambrogi asked that question in separate experiments on Tuesday. Colin ordered a detailed analysis of an evolving aspect of Ontario employment law, complete with advice to clients, and followed that by generating 30 termination clauses with varying degrees of immunity from judicial scrutiny and used them as a database to code an app for reviewing employment agreements. Bob requested (and received in ten minutes and 9,000 words) a memo on the legality of the Trump administration’s “temporary pause” of federal grant and financial assistance programs, and followed that up with a law practice management software research assignment.

You can assess all these results for yourself. But the takeaway for me is that Deep Research’s output shows extraordinary legal research and writing capacity. There’s just no getting around that this is lawyer work — mid-level associate work, to my eye, produced mere hours into Deep Research’s existence and, note this, without access to dedicated or proprietary legal databases.

And that’s just one new Gen AI legal capability. DeepSeek upended the AI and financial worlds and brought forth the prospect of open-source “Local AI.” Then there’s a demo I received last week of vLex’s AI-powered legal research product, which astonished me with its speed and accuracy. I’ve had to recalibrate my expectations for Gen AI in the legal space, and I’ve had to conclude: Generative AI is going to take over a lot of legal work, sooner than later. It’s going to eliminate lawyer jobs, reconfigure what lawyers do, and change the legal sector profoundly. We haven’t reached the far end of the spectrum, but we’re coming up on it with unnerving speed.

Many lawyers, understandably, react poorly to this idea. AI can’t be that transformational: It’s too early, it’s not proven, it still makes mistakes, what about the hallucinated caselaw? I get all that.

But what we don’t seem to notice is how we keep moving the goalposts: Well, okay, it can write a 9,000-word fully sourced legal memo on a critical constitutional issue in ten minutes; but are we sure every citation is correct? At a certain point, this starts to become skepticism for its own sake, waving off extraordinary developments because they feel like just too much to accept.

We could call this the complacency of continuity. Today is mostly similar to yesterday, which was mostly similar to the day before, and so on; therefore, tomorrow will be mostly the same as today. The legal sector, which has defeated almost all innovations and reform attempts over the last few decades, is especially prone to this. And most of the time, it works out fine. Most outcomes do normally fall between two extremes.

But not every time. Very occasionally, we encounter a period which is mostly extremes. “Weeks where decades happen,” as Lenin once said. We’re in one of those periods right now.

I’m sure you know only too well the thundering blitzkrieg of President Trump’s first two weeks in office — the illegal, the unconstitutional, the belligerent, the destructive, and the inhuman, day after day after day. It’s felt like a bad dream, and maybe that’s the problem. It seems so unreal and outrageous that people have trouble grasping what’s going on: This can’t actually be happening, can it? That’s the nefarious genius of taking what you want without permission or apology: Nobody can quite believe you’re doing it. We’re unaccustomed to the audacity of casual lawlessness.

That could be why Congress and the Democratic Party have been so slow to react to the dismantling of the rule of law. The cognitive dissonance generated by so many extreme events, all at once, has been too much. As I write, fear and anger are starting to overcome paralysis, and a protest movement appears nascent. Americans are starting to realize that extreme events are no longer outliers, but everyday reality.

Canadians, too. Our complacency ended when the Trump Administration announced 25% tariffs on all our exports, and it hasn’t returned even after the “30-day pause.” We didn’t want to believe something so destructive could actually happen; but we’ve had to accept that if Donald Trump really wanted to destroy our economy to make Canada part of the US, he could, and he would. The prospect of being hanged in a fortnight has now cleared a lot of minds here: We seem to be recognizing how we’re far more vulnerable than we’ve been willing to admit, and that our complacency is a big reason why. Hardly anyone here was talking or thinking this way even a week ago.

This isn’t a geopolitics newsletter, although events will likely compel my return to these topics in future. But this will be a rule of law newsletter, at least in part, as the times demand it. I’ll re-engage with legal industry commentary and analysis in due course.

But the point I want to leave you with today is that we can’t be complacent anymore, neither as citizens nor as lawyers. We can’t rely on old assumptions about how things are done, and about who does and does not have the power and the license to do them. I feel like we’re fast approaching a transformational period, in the legal sector and the wider world. And I don’t think we can avoid it. The only way out is through.

How can we get through? My default rule in a crisis is always to go back to first principles. As a practicing lawyer, focus on the people you serve and the values you hold; as a citizen lawyer, speak out for the rule of law and advance the materiality of justice. But in these circumstances, I think, we should also get ready to do something most lawyers really dislike: Be unreasonable.

From our first day in law school, we are immersed in reasonableness. It was held out to us as the highest virtue of the legally trained mind, and we’ve absorbed it as a core professional value. We are reasonable people, and we fully expect the world to behave reasonably, too. When it doesn’t — when lawbreakers reject norms, for instance, or when technology changes all the rules — we can become hesitant, unsure, and start to back off.

What we ought to do instead, I think, is to show the world what happens when lawyers stop being reasonable. An unreasonable response to Generative AI involves jettisoning all our assumptions about what lawyers do and staking a rightful claim to the newest and most valuable opportunities coming onstream. An unreasonable response to the dismantlement of democracy is to drop the gloves and leap into the fight with every weapon in the lawyer arsenal, to make our opponents understand that while we won’t normally start a fight, we sure know how to end one.

There’s a time to be reasonable, and there’s a time when we need to be fierce, combative, and disruptive, to advance our interests and protect our values. Spend some time today thinking about the power of unreasonable lawyers. Think about what you could accomplish with that power, in a time like this.