Are you watching the legal AI software data battles? Legal AI is a data-hungry beast. An AI model must be fed the text of tens of millions of court decisions, legislative documents, and sometimes other legal sources. That data is the new gold, and it’s why the battle over legal databases is heating up. A recent Canadian controversy illustrates the stakes. Caseway (the legal AI startup I co-founded) was accused of tapping into a publicly funded legal database (CanLII) to train our LLM model. This was nonsense, but it didn’t stop 30 media outlets from publishing about it when the lawsuit was filed.
The non-profit that hosts public court decisions sued the startup trying to innovate with court decisions. This was because while the law is public, CanLII claims we violated their terms of use by scraping data.
Monopolizing access to publicly available data
We were stunned. It felt like an attempt to guard their turf under flimsy pretenses. I even said in the media that this lawsuit is really about “monopolizing access to publicly available data under the guise of copyright and restrictive terms of use”. After all, taxpayers fund the courts, and the public has a right to the law. Yet when a technology company tries to make that law more accessible with AI, is it suddenly wrongful? The irony would be rich if it didn’t have such a chilling effect on legal innovation.
Our response at Caseway was unapologetic and bold. If CanLII wanted to play hardball, we’d bypass them entirely. We announced plans to build our comprehensive court decision database, effectively making CanLII “irrelevant” if that’s what it took. (Go big or go home, right?)
The message was clear: you can’t stop the AI revolution by hoarding data. And we’re not alone in this sentiment. Around the world, legal technology players are rushing to assemble proprietary troves of legal text to train their models. It’s no coincidence that Thomson Reuters paid $650 million to acquire Casetext, a startup at the forefront of legal artificial intelligence.
Casetext had early access to GPT-4 and used it to launch CoCounsel in 2023. This AI legal assistant can whip up research memos, contract analyses and more in minutes. Why did a giant like Thomson Reuters shell out so much? Because they recognize that the real power lies in combining advanced AI with domain-specific data. Thomson Reuters owns Westlaw (a massive legal research database); by grabbing Casetext, they’re effectively marrying deep legal data with cutting-edge AI expertise. It’s a love story of data and algorithms, poised to redefine how lawyers get information.
CanLII v. Caseway
Here’s the dirty secret: General artificial intelligence like ChatGPT has read much of the internet, but much of the law is not freely available on the open web. It lives in court archives, specialized databases, and PDFS that aren’t crawler-friendly. So, the best legal LLMS need access to data your average chatbot never saw. Companies and even research groups are scrambling to collect and curate legal texts. Some partner with courts or governments; others, like in our case, test the limits of “public” data by scraping or negotiating access.
This raises huge questions: Who owns the law’s text? Can a publisher or a non-profit institute claim exclusive rights over public legal documents? Those debates are now playing out in court, and the outcomes will shape the future of AI in law. IP and copyright fights, like CanLII v. Caseway, are just the opening act.
AI that masters legal data
It’s also worth noting that assembling this data isn’t trivial. Courts don’t exactly hand out neatly organized spreadsheets of their decisions. Often it’s like getting boxes of unindexed documents. Much technical heavy lifting goes into cleaning and structuring legal data. Someone must add all the metadata once you obtain court rulings (often in raw Word or PDF format). This includes case names, dates, and citations.
Organizations like CanLII and others have spent years doing this work. We don’t use any of their work product, and never have. The winners in legal artificial intelligence will be those who do a good job of dealing with both the technical and legal challenges of gathering this goldmine of text by negotiating access, paying for licenses, or fighting for the principle that the law should be free.
Technical and Ethical Minefields: Beyond the Data Grab
Building a legal LLM isn’t just a data challenge. It is also a technical and ethical minefield. Let’s start with accuracy. When an AI is churning out movie recommendations or helping draft marketing copy, a mistake is usually low stakes. Mistakes can mean a wrongful conviction, a lost child custody case, or a multi-million dollar liability. The tolerance for error is zero. The “hallucination” problem with general LLMS is hazardous in our field. A fake citation or a misquoted statute isn’t just an “oops”; it’s malpractice territory.
The legal community is waking up to this. In July 2024, the American Bar Association went so far as to issue guidance (Formal Opinion 512) saying that lawyers must understand the benefits and risks of AI and verify AI-generated outputs. The ABA reminded lawyers that if an AI tool spits out a draft, the lawyer better double-check every line. If Chatgpt makes up a legal case and you fall for it, that’s on you.
Confidentiality and privacy
Another challenge is confidentiality and privacy. Lawyers deal with ultra-sensitive information. Client secrets, personal data, trade secrets, you name it. Many lawyers (quite reasonably) freak out at plugging confidential memos or evidence into a third-party AI system. And they should be cautious: if you’re using a public tool like the free ChatGPT, whatever you input could theoretically be used to train the model further or be exposed (OpenAI says they have policies, but do you really want to risk it?).
Legal LLMS offer a way forward by being deployable in more controlled environments. Law firms opt for self-hosted or private cloud models, where the data stays within their governance. In other words, the AI can come behind the law firm’s firewall. This alleviates the fear that using artificial intelligence means leaking client info.
Many law firms prefer domain-specific solutions over general ones because they can demand privacy guarantees. The ethical duty to protect client confidence (Rule 1.6 for the ethics nerds) means any AI that can’t promise security is a non-starter. The good news is that legal LLMs can be used to maintain confidentiality with the right setups, but this requires investment in the proper infrastructure and vendor relationships.
We also have to consider bias and fairness. If a legal AI is trained on past cases, will it inherit the biases in those cases? Almost certainly, if we’re not careful. The legal justice system has historical biases against certain groups or favours certain kinds of litigants.
Reduce costs and improve accuracy
A naïvely trained model might internalize those and, say, be less likely to predict success for a claimant from a marginalized group because historically those claimants have succeeded less. This is a thorny issue. Ethical deployment of legal LLMs means we must actively monitor and correct biases, perhaps by curating training data or applying algorithmic fairness techniques. It’s doable, but it can’t be an afterthought.
There’s also the irony that not using advanced software can be unethical. If artificial intelligence can drastically reduce costs and improve accuracy, could it eventually be considered unethical not to use it in certain situations? We’re not there yet, but consider access to justice (which I’ll get to in a moment).
If refusing technology means you serve fewer clients or take longer to solve a legal problem, at what point does clinging to the old ways breach your duty of diligence or make your services unaffordable?
It’s a question the profession will have to grapple with. Judges have even started hinting that lawyers must adapt. In a recent panel, judges urged lawyers to embrace AI, noting that “it’s a must-do to survive in law.” When the folks in judge robes, who are usually the most conservative about change, are telling you to get on the AI train, you know it’s serious.
Automated Law System Backed by Legal Data
Beyond ethics, let’s talk economics. This is the real driver of change in the legal industry. Legal LLMS are about to shake up the law business like nothing we’ve seen before. Law firms, huge ones, have long relied on a pyramid model: armies of junior associates grinding away at research, document review, and drafting, all billed by the hour, funnelling work upward to a few rainmaking partners.
This model is directly threatened by artificial intelligence automation. Why pay a first-year associate $300 an hour to sift through caselaw or due diligence documents when an AI can do 80% of that work in seconds?
A recent study benchmarked AI against junior lawyers for contract review and found the AI matched or exceeded the humans’ accuracy, completed the reviews in a fraction of the time, and did it at an estimated 99.97% cost reduction. You read that right: the AI was just as accurate at spotting legal issues, but insanely faster and cheaper. The researchers called it a “seismic shift” and concluded that “LLMS stand poised to disrupt the legal industry”. If that’s not an economic earthquake, I don’t know what is.
What does this mean for law firms? For one, the leverage model is in jeopardy. It’s not hard to envision that in a few years, a partner with a good legal AI will be able to handle matters that used to require a team of five associates. The mundane work, which includes researching legal cases, checking citations, drafting boilerplate sections of briefs or contracts, can be offloaded to the machine learning software. This slashes the number of billable hours needed to deliver quality work.
Flat fees or value-based billing
Forward-thinking law firms will adapt by changing how they bill (moving away from pure hours to flat fees or value-based billing) and training their people to work alongside AI (so those junior lawyers become AI-augmented analysts rather than document drones).
Law firms stuck in the old mindset will try to resist. This may include not adopting AI to profit more from the old ways. But law firm clients aren’t stupid. When Fortune 500 companies realize they can get a legal memo in 2 hours instead of 2 weeks, or that one lawyer with AI can finish a deal due diligence in a day rather than a month, they will demand that. Law firms that don’t deliver it will watch clients walk away to those that do. Clients should ask their law firm, “Do you use AI?” If the answer is no, they should fire the law firm and get a new one.
Adoption of legal artificial intelligence
We’re likely to see a reordering of the legal industry. Early adopters are going to gain a serious competitive edge. It might even trigger an “arms race” in law, where firm A’s use of AI forces firm B to jump in and keep up. Despite some inevitable pushback and perhaps even litigation to slow things down (there are challenges to AI use or attempts to ban it in court filings), the economic pressure is too great. As one legal AI study put it, resistance to LLM adoption is often just protectionism of the status quo, not a genuine concern about the tech’s validity.
The reality is that broad adoption of legal artificial intelligence will radically reduce the demand for entry-level drudge work. That doesn’t necessarily mean fewer lawyers overall, but lawyers must move up the value chain. Tomorrow’s junior lawyers won’t spend their days proofreading leases or compiling case bundles; AI will handle those tasks. Instead, those young lawyers will engage in complex analysis, strategy, and advocacy sooner. These are the things AI can’t do (yet). In a sense, the human roles will become more interesting as the rote busywork evaporates. However, fewer total hours will likely be billed, which has profound implications for law firm staffing and profitability.
Law Firms Can Be Replaced By Legal LLMS
Any law firm that tries to ignore this trend will get left in the dust. A recent survey of over 200 lawyers found that 78% of law firms aren’t using any AI yet, often due to hand-wringing over data security or simply the inertia of “this is how we’ve always done it”. That means only 1 in 5 firms have started to embrace this software.
This gap won’t last long. The 22% who are experimenting now will rapidly pull ahead in efficiency. The other 78% will face a stark choice: catch up or fade away. Even the old-guard institutions are sounding the alarm. The ABA’s 2024 Tech Report noted that firms slow to adopt AI may find themselves “falling behind competitors” who offer faster, cheaper services.
Indeed, one of the most significant risks for law firms today is not adopting AI. Those who resist risk losing clients to more tech-savvy firms and even struggling to attract new talent. Who wants to work at a firm that still uses fax machines when their peers use AI copilots to supercharge their practice?
The next generation of lawyers
Let’s talk about talent for a second. The next generation of lawyers grew up with technology. They’re not intimidated by AI; they’re intrigued by it. Law firms that position themselves as AI-friendly will draw the best young talent, who see an opportunity to do cutting-edge work.
Those who ban AI out of fear will seem like career dead ends to ambitious graduates. We’re already seeing a shift: young lawyers want to leverage AI to eliminate drudgery. If you tell them, “Sorry, we don’t do that here, just put in your 80 hours manually reviewing documents,” don’t be surprised when they lateral to a firm that does use AI. In short, embracing legal LLMS is about staying relevant and competitive as a business.
Legal Intelligence Platform with AI and Data Integration
Perhaps the most inspiring aspect of legal LLMS, and the one I’m most passionate about, is their potential to improve access to justice dramatically. For years (decades, really), we’ve wrung our hands over the “justice gap.” In the United States, studies consistently show that most people who need legal help don’t get it. According to a 2022 report, low-income Americans received no or inadequate legal help for 92% of their civil legal problems.
Read that again: 92% went essentially unserved. Legal aid organizations are heroic but perpetually overmatched. They only meet about 1% of the need (with pro bono lawyers, maybe scratching another 2%). We won’t be able to bridge this gap with the traditional approach; there will never be enough affordable human lawyers. This is where legal AI could be revolutionary. Dramatically lowering the cost of legal services can help make legal services available to people who today get nothing.
Imagine a near-future scenario where you have a legal question or a minor dispute. If your landlord refuses to return your deposit, you must fight a parking ticket or draft a simple will. You can’t afford a lawyer’s $300/hour rate for such a minor issue. Instead, you use an app or website powered by a legal LLM. You describe your situation in plain English.
The AI instantly analyzes your issue and asks relevant questions (like a lawyer would in an interview). It then provides guidance. Perhaps it’s a draft letter invoking the correct statutes and case law to send to your landlord, a step-by-step instruction on filing that small claims case, or a properly formatted will you can execute.
Courts are swamped
Because it’s artificial intelligence, the cost can be pennies on the dollar compared to human help. We’re talking potentially $5 or $10 for something that might have cost $500 in legal fees, or free legal triage offered by courts and nonprofits using these models to help people help themselves.
The impact on access to justice could be enormous. Courts are swamped with self-represented litigants who don’t know the procedures or law, causing delays and often injustice. Legal AIS could guide pro se litigants through forms and filings, generating documents acceptable to the court.
Legal aid lawyers could multiply their impact by letting the AI handle routine drafting so they can focus on court appearances and negotiations for clients. Consider administrative tribunals (like benefits claims, immigration, and small tribunals). Many people give up on legal claims because the process is too complex. An AI assistant could change that by walking someone through it step by step at any hour of the day.
Artificial intelligence legal assistance
Now, I’m not naive; there are challenges here, too. We have to ensure the AI’s advice is correct and doesn’t lead someone astray in a way that hurts them. We likely need some regulatory sandbox to allow “artificial intelligence legal assistance” for the public without running afoul of unauthorized practice of law rules. (The first time an AI gives bad advice to a pro se litigant, you can bet someone will try to sue or regulate.)
Overall, I firmly believe legal LLMS can democratize legal knowledge. Access to basic legal information and help should not be the privilege of the few who can afford $400/hour lawyers. If the law is public and AI can interpret it for people at scale, we have a moral imperative to pursue that. This is why attempts to lock down legal data bother me. They could deny people a lifeline that technology could provide at virtually no cost.
I’ve long said, “Legal justice delayed is justice denied.” Well, justice unobtainable is the ultimate denial. Legal AI won’t fix systemic legal assistance underfunding or eliminate all injustices, but it can empower individuals in unprecedented ways. A well-designed legal LLM can be like having a knowledgeable legal aide in your pocket, ready to help immediately.
That could radically change the power dynamics between ordinary people and big institutions (landlords, debt collectors, bureaucracies) that bank on the fact that you don’t know your rights. When people can quickly understand “what my legal options are” without paying a week’s wages, that’s a huge win for society. It’s tech for good, in the truest sense.
Adapt or Die: Firms That Resist AI Are Falling Behind
Despite all these advantages, corners of the legal world are still digging in their heels. I’ve heard the arguments: “Our law firm has never needed this before, why start now?”, “Our lawyers are top-notch, we don’t trust a machine”, “What if the AI makes a mistake? We’ll just avoid it.” These excuses are starting to sound like the death rattle of a bygone era. History has not been kind to industries that resist innovation, and the legal sector will be no exception.
As noted in the Embroker article (sources at the end of this article), only about 22% of firms say they’re actively using artificial intelligence software so far, which means the majority are still standing on the sidelines. But that majority is shrinking by the day. We see announcements of firms piloting AI or partnering with legal tech companies every week. Those who continue to stick their head in the sand will wake up to find the world has moved on without them.
Legal AI Software and Data
The evidence is already here. One of the world’s largest law firms, Allen & Overy, deployed a GPT-powered legal chatbot (Harvey) to 3,500 lawyers across 43 offices and saw 40,000 queries submitted in the pilot alone. According to the firm, 25% of their lawyers used the AI daily, and 80% used it at least monthly.
That’s at a white-shoe, elite firm known for high-end work. If AI were going to be relegated to low-end tasks, you wouldn’t expect those adoption numbers in that environment, but there they are. And Allen & Overy isn’t alone; an estimated 28% of top US law firms already use Harvey or similar AI software in some capacity. The tide has turned. AI in law went from a curiosity to a mainstream productivity tool in about a year. This “rise of the machines” isn’t theoretical; it’s happening in the most traditional law offices.
Many law firms are falling behind because of AI
So what happens to the firms that don’t adapt? Simply put, they fall behind. They lose clients because clients will flock to firms that deliver faster results at a lower cost. They lose good lawyers, who won’t want to grind away like it’s 1985 when they could practice at a place embracing 2025.
As discussed in the New York State Bar Association source article, refusing to use available technology could become an ethical issue, even with efficiency aside. Some state bars and judges already hint that not staying abreast of technology could violate a lawyer’s duty of competence.
Think about that: you could actually be seen as an incompetent lawyer if you willfully ignore software that makes you more effective. Meanwhile, those who do adopt AI are poised for an “arms race” advantage: as they handle more volume or deliver work quicker, they’ll gobble up market share.
There’s a telling quote from an AI risk management report: “One of the biggest risks for law firms is not adopting artificial intelligence.” Law firms that resist will “likely face increased inefficiencies and difficulty attracting and retaining talent”. In other words, playing it safe is the more dangerous route. This flips the script on the traditional conservative mindset of law firms.
For once, being too cautious could cost you dearly. It reminds me of the early days of the internet. Some law firms refused to use email (for confidentiality reasons) or stuck to law libraries, while others embraced online research. The law firms that moved with the technology ended up leaps and bounds ahead. The same thing is starting now with AI. A managing partner who says “we’ll never use AI here” today effectively declares that they intend to stagnate. Clients will notice, as will the competition.
Will lawyers be replaced?
A lot of the resistance is not due to unproven technology; it’s due to the threat to the law firm business model. Some law firm leaders are dragging their feet because they worry about how we will bill if artificial intelligence makes the law firm run too well. When you’ve built an empire billing by the hour, a tool that cuts hours by 70% seems scary. In a survey, many law firms admitted that the billable hour model diminishes the urgency to innovate.But clinging to an outdated billing model is a recipe for disruption.
Rather than fight the future, savvy firms are reimagining pricing and value. Clients hate the billable hour anyway. They want results, not time. AI allows progressive firms to deliver better results faster and charge for the value delivered. Those who insist on charging for inefficiency (“we refuse to use AI so we can bill you more hours”) invite clients to fire them. Resisting technology to protect billables is a short-sighted strategy that will backfire as the market evolves.
-Alistair Vigier
