She enrolled in an online Juris Doctor program instead. But she also kept her day job, and clerked remotely for a law firm she has still never visited in person.

Her path would have been unrecognizable to the generation of lawyers before her. For decades, the route into the profession ran through a single, rigid channel… A campus law school, an articling position secured through proximity and connections, and a working life built around a physical office and the assumption that everyone would be in it.

That channel still exists. It is no longer the only one.

Remote work, digital collaboration tools and shifting expectations about where work happens have reshaped how lawyers are trained and how they practise. What began as an emergency measure during the pandemic has hardened into something more permanent. It has reached law schools, employers and the firms that hire new graduates.

The result is a profession that increasingly looks nothing like the one earlier generations entered.

Pursue a Juris Doctor without abandoning a salary

The change starts with legal education itself. Aspiring lawyers are weighing cost, access and flexibility more heavily than ever when deciding how to earn a degree. Institutions have responded by expanding hybrid and fully online programs. A working professional with family obligations can now pursue a Juris Doctor without abandoning a salary or relocating a household.

According to Cleveland State University, some of these programs deliver coursework entirely online while preserving in-person contact through residential weekends. It’s a hybrid model designed to keep students connected without tethering them to a campus.

The shift in education tracks a broader change in the workplace. The International Bar Association has described the emergence of what it calls the “hybrid legal function.” It is where in-house counsel, alternative legal service providers and technology-driven systems operate together under far more flexible arrangements than the profession once tolerated. Cost pressure is driving much of it… In one IBA survey, roughly 42 per cent of law firms said they had been asked to cut costs over the prior year.

The geography of early-career work has loosened too. Internships and clerkships were once gated by location. It was only available to students who already lived in a major legal market… Or those who could afford to relocate for a summer. Many organizations now offer remote or partially remote placements. They let students contribute to research, contract review and case preparation from anywhere. Law firms that were once unreachable for a student in a small city are now a video call away.

Become a lawyer remotely

Technology has moved from the margins of legal training to its centre. Cloud document systems, e-discovery platforms, AI-assisted research and virtual hearings are now routine across much of the profession, and graduates are expected to understand not only legal doctrine but the systems that run a modern practice.

A Thomson Reuters report on in-house lawyers found that hybrid arrangements are now treated as a long-term reality rather than a temporary accommodation, while flagging the confidentiality and collaboration risks that come with them.

There is just one problem: the model is far from solved.

Mentorship is the sharpest concern. Junior lawyers have always learned by watching, by the overheard hallway conversation and the unscripted minutes after a meeting. Those moments are hard to manufacture over video. Research on remote onboarding suggests that employees who start their careers entirely at a distance can struggle to build attachment to an organization, raising the odds they leave early.

The risks are practical as well. Lawyers handle confidential material, and distributed teams strain both cybersecurity and the simple consistency of communication across time zones.

Even so, the profession keeps adapting. Many law firms are settling on hybrid structures that pair remote flexibility with periodic in-person collaboration, an attempt to keep the access without losing the apprenticeship.

The path into law is no longer tied to a single model of where to study or where to work. What looked like a temporary disruption has become a lasting reshaping of how lawyers are made.

The post How to Become a Lawyer When the Office Is Optional appeared first on Advocate Daily.

Selection is free, open application, and judged by an outside panel. The 2026 Vancouver jury included Klue co-founder Jason Smith, Michael Lints of Golden Gate Ventures, Jen Holmstrom of Notable Capital, and Bigbang Angels’ Byung Sun Hwang, among others.

The 2026 cohort is good company. Hydra Energy, MetaOptima, Vsblty, LifeBooster, HTuO Biosciences, Butterfly Legal, and Divvi Health are all on it. Maninder Dhaliwal chairs Scale Up Canada, a not-for-profit. Its jury applied three straightforward criteria. Canadian innovation with global potential, measurable positive impact, and a contribution to the local economy. Caseway clears all three. We measure ourselves by one test: does what we build help people the legal system has priced out?

A quick word on the path here, because it was not a straight one.

The Story Behind Caseway

Seven years in the Canadian Army, including JTF2 pre-selection, taught me how to decide under incomplete information and live with the result. Five years inside divorce law taught me what the system actually feels like for the people inside it: expensive, slow, and indifferent to anyone who cannot afford $400 an hour to be heard. Ten years on the board of The Last Post Fund taught me what showing up for people after the institutions have moved on looks like. All three observations are in the product.

Before Caseway, there were two other companies. Jusu is a plant-based wellness brand with roughly 60,000 customers, which I reacquired in 2025 and still operate.

Clearway became the largest lawyer review site in Canada and the UK, with around 200,000 monthly users at peak. Clearway proved that consumer-facing legal information scales as a distribution problem more than a content problem. Caseway answers a different question: what does the legal workflow look like if research and paperwork are no longer done by a human?

Scale Up Canada 2026 Vancouver50

That is the bet. Casey handles research and drafting. CaseForm automates court and regulatory forms and is now live inside MyCase via 8am and AffiniPay. Synthium is the enterprise version of the same data brain, industry-agnostic by design and already moving outside legal.

The work is being validated where it should be. UBC’s Dr. Vered Shwartz and her group are running a two-year NSERC-funded study on our retrieval engine.

SFU’s School of Computing Science announced a parallel collaboration in January on whether searchable case law improves outcomes for self-represented litigants.

Business in Vancouver covered why our hallucination rate is where it is.

Our integrations and partnerships are listed publicly.

Vancouver50 is a marker, not a finish. Vancouver has produced Clio, AbCellera, Trulioo, Klue, and Visier in roughly the same window. The bar is set and we intend to clear it.

The post Caseway named to the Scale Up Canada 2026 Vancouver50 appeared first on Advocate Daily.

The agreement marks both companies’ formal entry into dual-use defence technology, with an explicit focus on the protection of Canada and the broader NATO alliance.

The agreement marks both companies’ formal entry into the dual-use defence technology space, with an explicit focus on the protection of Canada and the broader NATO alliance.

First project: True North Precision

The partnership’s first joint pursuit is True North Precision. It is a bid into the Department of National Defence’s IDEaS Competitive Project, under the MINERVA initiative. True North Precision delivers low-cost ISR drones with eye-safe laser rangefinding and target cueing.

The system is engineered to support indirect fire and battlefield awareness at the platoon and company level. It feeds structured target data and 10-figure MGRS geolocation directly into ATAK and existing C2 workflows.

ITSP.10.171-compliant information handling is baked in from day one. Caseway acts as prime contractor on the Canadian bid. It leverages its NCAGE registration, IRAP engagement, CanadaBuys profile, and prior successful contract execution history.

Valtec leads platform and rangefinder integration. Final Capability Assessment is targeted for May 2027 at TRL 7.

A bilateral, jurisdiction-aware partnership

The two companies have structured the partnership to pursue opportunities across allied jurisdictions. Caseway primes Canadian government bids out of Vancouver. Valtec primes United States bids out of its Delaware footprint. Multilateral and NATO-aligned opportunities will be jointly prosecuted based on best fit. The partners are exclusive to each other on integrated detection-to-decision solutions for defence customers in agreed jurisdictions.

The pipeline beyond True North Precision includes joint Canada-US bids and NATO-aligned opportunities where allied interoperability, GNSS-resilient operation, and audit-grade data handling are non-negotiable.

Canada defence tech partnership: quotes

Alistair Vigier, Founder and CEO, Caseway: “Detection without decision is just data. The Caseway-Valtec partnership closes that loop. We are bringing structured, audit-ready outputs to the edge of the battlefield, and we are doing it with a partner whose hardware is genuinely best-in-class. Canada and our NATO allies need this capability now.”

John Keh, Founder and CEO, Valtec: “Valtec builds platforms that perform in the conditions operators actually face, contested GNSS, weather, weight constraints, and short timelines. Pairing that with Caseway’s decision layer means a soldier on the ground gets a target solution that a commander can act on, with the provenance to back it. This partnership is built for the threats Canada and its allies face today, not five years from now.”

Academic foundations

Caseway’s decision and data layer is developed in active research collaboration with Simon Fraser University and the University of British Columbia, providing independent academic validation.

See:

SFU: https://www.sfu.ca/fas/computing/news-events/news/2026/january/sfu-launches-legal-ai-collaboration-with-caseway-to-improve-acce.html

UBC (Dr. Vered Shwartz): https://www.cs.ubc.ca/news/2025/12/dr-vered-shwartz-collaborates-vancouver-based-tech-startup-study-legal-research-engine

About Caseway

Caseway is a Vancouver-based AI company building structured, audit-ready decision and data infrastructure for regulated and high-stakes environments. Caseway is engaged with NRC IRAP, registered on CanadaBuys, NCAGE-registered, and an IDEaS participant. More at https://www.caseway.ai

About Valtec

Valtec is a defence-technology company with a US Delaware footprint, building compact, GNSS-resilient unmanned aerial systems for ISR and precision targeting missions. More at https://www.valtec.ai

Media contacts

Caseway | Alistair Vigier, Founder & CEO

Valtec | John Keh, Founder & CEO

The post Canada Defence Tech Partnership: Caseway and Valtec Join Forces appeared first on Advocate Daily.

Vexatious litigation in Canada and the United States is jamming up the courts. The systems used to identify these litigants are slow, reactive, and built for a paper era. The cost is a justice system that runs slower for everyone else.

In 2018, after six years of fighting a credit union over her late mother’s estate, after passing through the hands of 23 different judges, and after running up more than $243,000 in court-ordered costs against herself, Janet Hoessmann was finally banned from suing without permission. Justice Kenneth Ball of the BC Supreme Court called her conduct “reprehensible and remarkably wasteful.” That was the formal language. The plain version is that one self-represented litigant in one province burned through more judicial time than most full-blown commercial trials.

She is not unusual. She is just visible.

Vexatious Litigation in Canada: Nine Regimes, No Registry

In April of this year, Justice Nicholas Devlin of the Alberta Court of King’s Bench declared another woman vexatious after a five-year fight over a house sale. His closing paragraph reads less like a judgment than a piece of clinical prose: “The court sincerely hopes she is able to find relief from the obviously compulsive distress the sale continues to cause her, going on five years later. That relief will, however, not come from the court and must not come at any further expense in time, stress or resources to innocent parties.”

In Toronto, an Ontario lawyer named Shane O’Herlihy was declared vexatious last year for a campaign that included 392 emails and voicemails over 318 days, 2,374 cross-examination questions, and labelling a city employee a “Pride Killer.” A lawyer. Declared vexatious by his own court.

These are not edge cases. They are the surface of a much larger problem that the legal profession has been politely refusing to discuss for about forty years.

What the word actually means

The Canadian Bar Association’s National magazine recently ran a piece titled “The vexing problem of vexatious litigation.” The article quotes Donald Netolitzky, the former complex litigant management counsel at the Alberta Court of King’s Bench and the closest thing Canada has to a national authority on the topic. His framing is worth repeating because most lawyers I talk to get it wrong.

“Vexatious has a very simple commonplace meaning,” Netolitzky said. “It means that you’re a problem, you’re out to hurt somebody, you’re out to vex them.”

That is distinct from frivolous, which means a claim with no legal basis, and abusive, which means using the court process for an improper purpose. A litigant can be one, two, or all three at once. Most repeat offenders eventually become all three.

Netolitzky’s other observation is the one that should be driving policy: “This is really a mental health phenomenon, in my opinion, more than anything else. Courts are not intended to handle people’s mental health issues.”

His 2017 study of self-represented appellants at the Supreme Court of Canada found that approximately 70 percent had an abusive litigation record by the time they reached that level, and roughly 40 percent had identifiable mental health problems. Those are not numbers from a tabloid. That is the actual filtered population at the top of the country’s appellate pyramid.

The law on the books

Canada has nine different vexatious litigant regimes for nine different jurisdictions, plus a federal one. Section 40 of the Federal Courts Act lets the Federal Court bar persistent vexatious filers, but only with the consent of the Attorney General of Canada. Ontario uses section 140 of the Courts of Justice Act, which was significantly expanded in August 2024 under Bill 157 to let judges issue orders on their own initiative and to give the Court of Appeal jurisdiction it never previously had. British Columbia uses section 18 of the Supreme Court Act.

Alberta uses sections 23 and 23.1 of the Judicature Act, supplemented by the Hok two-step procedure that the Court of King’s Bench developed in 2016.

Quebec is the outlier. It runs an actual public registry under article 84 of the Code of Civil Procedure. As of 2022, the Montreal district list named 336 individuals. That is the only formal vexatious litigant registry in Canada. Everywhere else, you have to assemble the list yourself by searching CanLII case by case.

The American system is fragmented in a different way. There is no federal vexatious litigant statute. Federal courts rely on the All Writs Act, 28 U.S.C. section 1651, plus inherent supervisory power, plus Rule 11 sanctions. The leading test comes from the Ninth Circuit’s 1990 ruling in De Long v. Hennessey, which requires four things before a prefiling order can issue: notice, an adequate record, substantive findings of frivolousness, and an order narrowly tailored to the specific conduct. The Second Circuit added a five-factor version in Safir v. United States Lines.

How Vexatious Litigation in Canada Actually Works

State law is where it gets interesting. California’s Code of Civil Procedure section 391, first enacted in 1963 and most recently amended in 2022, is the gold standard. The Judicial Council of California maintains a public list that is updated monthly. The current PDF, dated April 1, 2026, runs to several thousand entries. Texas runs a similar list under Chapter 11 of its Civil Practice and Remedies Code.

Florida’s regime, originally enacted in 2000, was substantially expanded in July 2025 to extend the lookback period from five to seven years and to include family law, small claims, and out-of-state cases. Ohio, Hawaii, Nevada, and a handful of others have parallel statutes.

Quebec has 336 registered names and California has thousands. The two jurisdictions are roughly comparable in population. Either Quebeckers are dramatically better behaved, or one of these systems is actually finding the problem and the other is not.

The numbers nobody publishes

Canada does not produce a national statistic on vexatious litigation. Nobody does. What exists are proxies, and the proxies are unflattering.

Trevor Farrow’s Canadian Forum on Civil Justice cost-of-justice study, published out of Osgoode Hall Law School, found that Canadians spend $7.7 billion annually on everyday legal problems. The state spends another $799 million on the social costs that flow from unresolved disputes. The average Canadian spends $6,100 per legal problem, and 51 percent of people experiencing a problem report increased stress as a result. None of that money distinguishes between productive litigation and vexatious litigation, which is part of the problem.

Self-represented litigant rates tell a similar story. Julie Macfarlane’s 2013 National Self-Represented Litigants Project study found that 64 percent of family court cases in Ontario, and 74 percent of cases in Toronto’s two main family courts, were begun by people without lawyers. Macfarlane updated that figure in 2021: “In the 1980s, about one percent of people represented themselves in court.

The National Self-Represented Litigants Project

Today, in family courts, it’s over half.” The vast majority of self-represented litigants are not vexatious, but most vexatious litigants are self-represented. The two populations overlap heavily, which is why the National Self-Represented Litigants Project has been warning since 2019 about what it calls “vexatious lite,” meaning judges using the language of vexatiousness without ever issuing the formal order.

The American numbers are easier to find. The Administrative Office of the U.S. Courts reported that pro se filings made up 50 percent of new appellate cases in 2025, totalling 20,878 cases, up nine percent year over year. Civil cases pending more than three years rose 346 percent over two decades, from 18,280 in March 2004 to 81,617 in March 2024.

In Canada, criminal-court delays under R v. Jordan have produced an even starker number. According to Globe and Mail and Statistics Canada reporting, 9,560 cases were stayed or withdrawn for unreasonable delay in 2023 to 2024. The peak was 11,132 cases in 2022 to 2023. About 10,000 criminal prosecutions a year are getting tossed because the system cannot move them through fast enough. Chief Justice Richard Wagner of the Supreme Court of Canada was unusually direct about it at his June 2024 press conference: “One stay of proceedings is one too many, let’s be clear on that.”

Vexatious litigation is not the only reason for the backlog, but it is a meaningful slice of it that almost nobody quantifies.

Why we do not flag earlier

Netolitzky’s other criticism in the National Magazine piece is the one I keep coming back to. Canada’s vexatious litigant regime, he said, is “a backwards, punitive sort of process.” It only works after the damage is done. By the time a court declares someone vexatious, that person has typically already filed dozens of actions, generated hundreds of thousands of dollars in costs, and consumed hundreds of hours of judicial time.

The hallmarks are well known. Justice David Stratas of the Federal Court of Appeal laid them out in Olumide v. Canada in 2017. Repetitive relitigation. Frivolous appeals. Refusal to pay costs. Allegations of corruption against opposing counsel and judges. Personal attacks. Volume for the sake of volume. None of these are subtle. A clerk who has been on the job for six months can recognize them.

The State of Vexatious Litigation in Canada

The problem is that nobody is connecting the dots across jurisdictions. In National Bank of Canada v. Taha in 2022, the Federal Court declared a litigant vexatious only after the Prince Edward Island Supreme Court had already done so. The opposing parties had to re-prove the entire history because there is no automatic interprovincial recognition. The same person who is barred from filing in PEI can walk into a courthouse in Calgary the next day, and unless the prothonotary happens to recognize the name, nothing stops them.

California’s section 391 statute does at least allow out-of-state declarations to count as a basis for new findings. Texas does the same. But each state still maintains its own registry, and there is no national American registry either.

This is a data problem dressed up as a legal problem.

What pseudolaw looks like

A category of vexatious filing worth flagging on its own is what Associate Chief Justice John Rooke of the Alberta Court of King’s Bench labelled “Organized Pseudolegal Commercial Argument” or OPCA litigation, in his 188-paragraph 2012 ruling Meads v. Meads. That decision is now the most-cited Canadian judgment internationally on the subject of pseudolaw.

Rooke’s prose is bracing. “When reduced to their conceptual core, most OPCA concepts are contemptibly stupid.” And: “OPCA litigants appear, engage in a court drama that is more akin to a magic spell ritual than an actual legal proceeding, and wait to see if the court is entranced and compliant.”

Mr. Meads had served Justice Rooke a “fee schedule” demanding two million dollars in gold or silver every time the judge dismissed one of his arguments. Rooke noted dryly that the gurus who sell these strategies could not name a single instance in which the spells had ever worked. “Your spells, when cast, fail.”

OPCA is the most dramatic version of the phenomenon, but the underlying behaviour is the same as garden-variety vexatious litigation. Persistent filings. Belief that the court is corrupt. Claims that grow more grandiose with each loss. Complete refusal to engage with the law as it actually exists.

The American sideshow

The United States has the most prolific vexatious litigant in legal history, and his story is worth telling because it tells you something about how broken the prefiling-order system is when applied piecemeal.

Jonathan Lee Riches filed his first lawsuit on January 8, 2006, while incarcerated at the federal medical centre in Lexington, Kentucky. By January 2008, PACER showed 474 actions. By June 2008, the Western District of Wisconsin reported he had filed 1,834 lawsuits, 1,540 of them in the first six months of 2008 alone. Prison Legal News estimated his total at over 3,000 federal cases by 2013. Riches himself, in a lawsuit against Guinness World Records for naming him the most litigious person in history, claimed somewhere between four and five thousand.

His named defendants included Britney Spears, Steve Jobs, Bill Belichick, Janet Reno, Plato, Nostradamus, the Lincoln Memorial, the Eiffel Tower, the Holy Grail, Pluto, Three Mile Island, Plymouth Rock, the Roman Empire, Adolf Hitler’s National Socialist Party, the Olsen Twins, Kim Kardashian, and at one point 790 additional defendants spread across 57 pages.

In March 2008, Judge Willis B. Hunt Jr. of the Northern District of Georgia dismissed 276 of Riches’s cases in a single order. The federal courts eventually issued a permanent injunction. But the injunction came years and thousands of filings into the campaign, and only because each district had to independently arrive at the conclusion that he was a problem.

Compare that to the early-warning value if a single national database had flagged the second filing instead of the two thousandth.

Where the technology actually sits

A reasonable person reading the above would assume that some American or Canadian or British court has built an algorithm to flag this stuff. They have not.

I could not find a single court in any common-law jurisdiction that has publicly deployed an AI system to identify vexatious or repeat filers at intake. The closest active practice is procedural. Alberta’s Civil Practice Note 7, in force since 2018, gives judges an “Apparent Vexatious Application or Proceeding” summary procedure. Ontario’s August 2024 Regulation 322/24 lets a registrar refuse to file documents from someone already declared vexatious. Both are improvements. Neither is automation.

Pretrial criminal-risk assessment tools like COMPAS and the Public Safety Assessment exist, and they have been the subject of a decade of fierce litigation and academic critique, including ProPublica’s 2016 investigation and the Wisconsin Supreme Court’s State v. Loomis ruling. Whatever you think of those tools, they are not solving the vexatious litigation problem because they were not built to.

The Conflict Analytics Lab

Samuel Dahan at Queen’s Law runs the Conflict Analytics Lab, which has built OpenJustice and MyOpenCourt. OpenJustice now handles roughly 20,000 queries a month. The Law Foundation of Ontario gave the lab $467,000 to build an AI tool for Pro Bono Ontario’s housing work. Dahan’s framing is correct: “Ninety percent of cases are settled outside the courts and litigants are self-represented. We’re building an AI tribunal for small claims disputes.” That is the demand side of access to justice. None of it touches the courts’ ability to identify problem filers.

The reason is not technical. The reason is that Canadian and American courts do not have the data infrastructure to flag patterns in the first place. CanLII publishes only what judges send it, and only the published decisions, not the docket. Most provincial superior court dockets are not online or are online behind paywalls or searchable only one case at a time. The federal court and BC do reasonably well. Ontario’s new Public Portal launched in October 2025 in Toronto only and will not be province-wide until approximately 2030.

You cannot run pattern recognition on a system that does not publish patterns.

The conflict of interest in plain view

I have a stake in this. Caseway, the company I run, was sued by CanLII in November 2024 for accessing court data the public pays for. We settled in 2026. CanLII is a not-for-profit owned by the Federation of Law Societies of Canada, and it is currently the only place a Canadian can search judicial decisions for free. Its mandate is open access. Its policy on bulk data has been less open. Reasonable people can disagree about whether that is the right balance.

What is harder to disagree about is the consequence. If the people closest to the problem cannot get access to the data, the people running the courts will not be the first to spot the patterns. They will be the last. They have been the last for forty years.

Ade Olumide was declared vexatious in three separate forums (the Federal Court, the Federal Court of Appeal, and the Ontario Superior Court of Justice) and the Supreme Court of Canada denied him leave to appeal four times before the situation stabilized. Each declaration was issued independently. Each cost time and money to obtain. None of them communicated automatically with the others.

If a private credit reporting agency were running this system, it would be embarrassing. The justice system runs it because there is nothing else.

What good would look like

I think there are five things that should happen, and only one of them requires AI.

First, every Canadian province should publish its vexatious litigant orders the way Quebec does. Section 84 of the Quebec Code of Civil Procedure has been on the books since 2009. It is not radical. It is not technologically demanding. The other provinces have not done it because they have not been asked to.

Second, court dockets should be searchable. Not the decisions, the dockets. PACER is not a perfect system, but the United States can tell you in seconds how many federal cases a person has filed. Canada cannot, and it is not because we lack the technology.

Canada should also adopt the equivalent of California’s section 391(b)(4) recognition rule, so that an Ontario vexatious litigant order is automatically treated as a relevant factor in BC. The provinces already recognize each other’s commercial judgments. Vexatious orders should be no harder.

Fourth, the courts should use the data they already have to flag patterns at intake. A clerk in Edmonton should not need to remember a name. The system should remember it.

Vexatious litigation in Canada

Fifth, and only fifth, AI can help. Not as a judge or as a gatekeeper, but as software that helps overworked clerks notice when they are seeing the seventh filing this year from someone they should be paying attention to. The same technology that lets a self-represented tenant find a useful precedent for her residential tenancy hearing can also help a court spot the person who filed 392 emails about the same dispute.

Justice Stratas put the underlying principle as well as anyone in Olumide. The courts, he wrote, are “community property” and “scarce natural resources.” They exist “to further access to justice by those seeking the resources of the court in a proper way.”

That principle cuts in both directions. We are not going to solve the access-to-justice problem by being afraid to identify the small minority of users who are consuming a disproportionate share of the resource. And we are not going to solve it by treating every self-represented litigant as if they were that minority either.

The most useful test, in my experience, is the one Justice Devlin reached for in his April ruling. Compulsive distress is not a legal category. But it is recognizable, and a system that can recognize it earlier will be a system that protects both the vexatious litigant from herself and the rest of us from her.

We are not there yet. We could be.

_______________

Alistair Vigier is the CEO of Caseway, a Vancouver-based legal AI company. He previously worked in divorce law, and served seven years in the Canadian Army.

The post Vexatious Litigation Canada: The Data Problem Nobody Is Solving appeared first on Advocate Daily.

The CBA just published an article titled “The vexing problem of vexatious litigation.”

That gap is what a Vancouver company called Caseway was built to close.

Canada’s legal data desert

Founded by Alistair Vigier, who served seven years in the Canadian Army and spent five years in divorce law before turning to technology, Caseway argues that the path to a fairer justice system runs through better data. The path runs through the hundreds of thousands of judicial decisions already produced by Canadian courts and scattered across inconsistent, fragmented, often inaccessible publishing systems.

“Judges are paid by the public to write these decisions. The public should be able to read them, search them, and learn from them,” Mr. Vigier said. “Right now, most people can’t.”

Canada’s court records problem has become a familiar refrain among legal technologists, academics, and access-to-justice advocates. About half of Canada’s provinces and territories have no online court search capability at all. In the jurisdictions that do, portals often confirm a record exists without giving the public access to the record itself. Verdicts delivered orally rarely make it onto paper, let alone online. Courts decide individually whether their written decisions are sent to CanLII, the not-for-profit that hosts the country’s free database of rulings.

The result is a country where the principle of precedent, the foundation of common law, operates on incomplete information.

Other countries have moved faster. The United States operates PACER, a federal database holding more than a billion documents. Harvard runs the Caselaw Access Project. France has Judilibre. Britain has the Cambridge Law Corpus. Canada has no equivalent.

A Vancouver bet on automation

Caseway’s product line was designed for that environment. The flagship, Casey, is a research and drafting assistant that pulls from publicly available case law to help lawyers and self-represented litigants quickly locate relevant authority, draft submissions, and stress-test legal arguments.

CaseForm automates the legal forms that bog down small firms, integrating with existing practice management software so partners can spend less time on intake and more time on advocacy. Synthium, the company’s enterprise product, brings the same automation logic to large institutions handling high volumes of structured data.

The common thread, Mr. Vigier said, is treating publicly created information as a public resource.

“We are an automation company. The legal market just happens to be the one with the most broken data infrastructure,” he said. “If you can fix the data, you can fix a lot of the cost problem at the same time.”

Legal fees

The cost problem is real. Canadians routinely report that legal fees price them out of disputes worth pursuing. In landlord and tenant matters, immigration appeals, small claims, and family law, self-representation has become the norm rather than the exception. Family law in particular is a domain where one or both parties almost always show up unrepresented, a pattern Mr. Vigier watched develop during his years practising in Vancouver. The gap between the people who need legal help and those who can afford it continues to widen.

The company operates exclusively through partnerships on a revenue-share model, working with bar associations, legal publishers, and software providers rather than selling directly to law firms. Recent collaborations include a research partnership with Simon Fraser University and a joint project with University of British Columbia computer science professor Vered Shwartz on the reliability of AI-generated legal answers. Caseway has also written publicly about hallucinations in legal AI in Business in Vancouver.

The “Lawfia” fight

That candour has not always been welcomed. In November, 2024, CanLII sued Caseway for copyright infringement, alleging the company had improperly accessed its database. Mr. Vigier, who previously called the lawsuit emblematic of what he describes as a “Lawfia” mentality in Canadian legal data, said the matter has since been resolved. He has used the dispute to make a broader point.

“You cannot claim proprietary rights over the work of judges, paid by the taxpayer, and then tell the public they need permission to learn from it,” he said. “That is the question this country has not answered.”

Critics of the AI-in-law movement have raised legitimate concerns about hallucinations, confidentiality, and the risk that automation entrenches existing inequalities rather than dismantling them. Mr. Vigier said those concerns are why Caseway invests in retrieval grounded in verified sources, partners with academic researchers on accuracy testing, and publishes openly about the limits of the technology.

He also pushes back on the framing that AI threatens lawyers.

“Lawyers who use these tools well will outcompete lawyers who don’t. That is the only real story here,” he said. “The clients still need a lawyer. They just need one whose hourly rate reflects the actual time spent thinking, not the time spent searching.”

Canadian Legal AI: Who actually benefits

For self-represented litigants, the calculus is different. A tenant facing eviction, a small-business owner responding to a demand letter, a parent navigating a custody dispute. None of them can afford to pay a research lawyer to find five comparable cases. AI can do that work in seconds, at a fraction of the cost.

That, Mr. Vigier argues, is what democratizing the law actually looks like in practice.

“The legal system was never supposed to be a vault,” he said. “Caseway is trying to open it back up.”

The post How a Vancouver company is trying to fix Canada’s legal data desert appeared first on Advocate Daily.

A Saskatchewan judge issued a ruling last July that one lawyer called groundbreaking. It dealt with whether police officers accused of wrongdoing can investigate themselves. Six months later, the ruling still isn’t on CanLII. The public can’t read it. Other lawyers can’t easily cite it. For practical purposes, it might as well not exist.

That’s the state of court transparency in Canada in 2026. And while courts on both sides of the border argue about who decides what gets published, artificial intelligence is transforming legal practice anyway. The question for court leaders now is simple: are you going to shape how AI enters the justice system, or are you going to let it happen to you?

I run Caseway, a Vancouver company that reads millions of court decisions and helps lawyers and self-represented litigants navigate them. We were sued by CanLII in late 2024 for building tools on top of Canadian court data. The lawsuit has since been resolved. But the experience taught me something I want court leaders in Canada and the United States to understand… The rules of the road for legal AI are being written right now, in courtrooms and legislatures, by people who may or may not know what they’re building.

The US is already further ahead, and the gap is growing

In the United States, PACER provides access to over a billion federal court documents. Harvard Law School built the Caselaw Access Project. Researchers, founders and journalists can actually study the system they operate in. In Canada, about half the provinces and territories have no online court search at all. British Columbia, Manitoba and Nova Scotia require a commercial business like mine to seek court approval before deploying AI tools on its database. That’s not caution. That’s a moat.

I’m not arguing the American model is perfect. It isn’t. But the gap matters because AI runs on data. If courts in your jurisdiction don’t publish decisions in bulk, machine-readable form, you are guaranteeing that the AI tools used by lawyers in your courts will be trained on data from somewhere else. Your case law, precedents, and your jurisdiction’s nuances. Missing.

Countries around the world are moving the other way. France has Judilibre. Britain has the Cambridge Law Corpus. These platforms are fairly standard in Europe now. Canada has a legal data desert.

The risks court leaders can’t delegate

I’ve spent the last two years talking to bar associations, regulators, academics and enterprise clients about AI in law. Here’s what I tell them, and it applies directly to the bench and the court administrators who support it.

AI doesn’t “know” anything. It predicts. Large language models are pattern recognition at scale. The risk isn’t that they’re wrong. The risk is that they sound confident when they’re wrong. Lawyers have already been sanctioned in both countries for filing briefs citing cases that don’t exist. That problem is solvable with the right data pipeline, but only if courts cooperate. Business in Vancouver wrote about how Vancouver AI firms are tackling hallucinations.

If your court publishes selectively, you are privatizing the law. Decisions of precedential value should be easy for anyone to find. When judges alone decide what gets published, the body of law becomes a curated collection rather than a complete record. Academics can’t measure whether like cases are decided alike. Self-represented litigants can’t see how similar disputes have resolved. Small law firms can’t compete with the firms that can afford Westlaw and LexisNexis subscriptions. Justice becomes a function of who can pay.

You should also know that the academic community wants to help. Caseway has research partnerships with Dr. Vered Shwartz at UBC, SFU’s faculty of computing, Northeastern University, and the University of the Fraser Valley. These aren’t consulting relationships. They’re researchers who want to study how legal reasoning works and whether the system delivers on its promises. Court leaders should be inviting that scrutiny, not hiding from it.

A practical starting point

If I were advising a chief justice or court administrator today, I’d say this. Pick one workflow, not the whole system. Start with something boring. Timely posting of written decisions. Machine-readable formats for appellate rulings. A clear policy on when oral decisions get transcribed. Put a human in the loop for review. Make one person accountable for the outcome.

Then measure. Count the decisions published within thirty days of release. You should count the decisions still sitting in a judge’s queue six months later. Have a look at the jurisdictions where a litigant can actually find the record of their own case online. If you can’t measure it, you can’t improve it.

Court Leaders – The real choice

AI is a force multiplier. It is going to make lawyers faster, make research cheaper, and make it possible for regular people to understand what’s happening in their own cases. That is happening whether or not courts cooperate. The only question is whether the systems being built on top of the law will be trained on complete, authoritative data or on whatever scraps developers can assemble elsewhere.

Courts exist to serve the public. Judges are paid by taxpayers to produce decisions that shape how the country is governed. Those decisions belong to the public. Not to vendors, databases operating at their own pace, or to individual judges choosing whether to hit publish.

Court leaders in Canada and the United States have a decision to make in 2026. Lead on this, or get led by it.

Alistair Vigier is the CEO of Caseway, a Vancouver-based legal AI company. He previously worked in divorce law, and served seven years in the Canadian Army.

The post A Wake-Up Call for Court Leaders in Canada and the United States appeared first on Advocate Daily.

At the same time, courts have made it clear (often through rejection) that “close enough” is not a strategy. One example with unusually transparent reporting is the Superior Court of California, County of Los Angeles eFiling rejection report: the overall rejection rate runs around 8–9% for civil filings and about 17% for family law filings (May through April in the report’s period), with “missing information or attachments” as the single largest family-law rejection reason. 

This is why “generic AI” is the wrong mental model for court forms. Large language models can be helpful for drafting and summarizing, but even legal-oriented models can hallucinate, and ethics authorities explicitly require careful human review and supervision when lawyers use generative software. 

The practical approach in this cycle is form-specific automation connected to the system where case truth already lives (your practice management platform). CaseForm’s direct integration with MyCase is built around that idea that you should reuse already-entered matter data, apply jurisdiction-aware rules, run validation checks, and keep an audit-friendly workflow inside your existing operations. 

Thesis and why this year matters

The best way to automate court forms right now is to treat them as structured compliance outputs, not “documents to draft.” That means the winning stack looks like using utructured case data in practice management → rules and validation tailored to each form and jurisdiction → fast generation → human review → filing and status tracking. Anything else is either

(a) a copy-paste treadmill, or

(b) a risk-transfer mechanism you will still be liable for. 

This matters more now because court operations are already heavily digital, so the bottleneck has moved upstream. The American Bar Association reports that 85% of litigators are already using electronic court filings and 73% of law firms use cloud-based legal software, meaning most firms have already crossed the “digital filing is normal” threshold. 

The adoption of AI software is rising, but trust and governance are lagging. The ABA’s TechReport analysis of its survey data puts current office AI-software usage around 30% overall, and reports that accuracy is the dominant concern (roughly three-quarters of respondents flagged it).  The Thomson Reuters professional services research similarly shows fast growth in genAI use in legal (from 14% to 26% year over year in its report), while also noting that many organizations lack policies. 

So the thesis is not “use more AI.” It is “stop wasting skilled humans on preventable transcription and preventable rejections.”

The data-driven cost of manual court forms

Manual court forms create three compounding costs: labor drag, rework from rejection, and opportunity loss from delayed or constrained throughput.

Start with volume. The Administrative Office of the United States Courts reports U.S. federal district civil filings of 347,991 for the year ending March, with civil filings up 22% year over year. Regardless of whether your firm touches federal court, the direction is plain: the system is not getting simpler or less paper-heavy, it is getting more document-intensive. 

Then consider rejection risk. In Los Angeles Superior Court’s public rejection reporting, civil efilings show an overall rejection percentage hovering around 8–9%, and family law around 17% in the same reporting window.  The top family-law rejection reason is “missing information or attachments,” often accounting for roughly a third of rejected submissions in the report.  That is not an edge case. It is a workflow failure mode.

Put a conservative price tag on the labor. The U.S. Bureau of Labor Statistics lists median pay for paralegals and legal assistants at $29.33/hour ($61,010 annually, May data).  If your typical court-form packet takes 60–90 minutes of staff time once you include finding the right form set, transcribing matter data, formatting, checking attachments, and packaging for filing, you’re looking at roughly $30–$45 in direct staff cost per packet before attorney review and before rework. 

The uncomfortable truth: if you are “solving” this by adding headcount, you are buying expensive capacity to do unscalable work. And you will still have the same rejection dynamics.

Why generic AI and legacy automation fail in practice

Generic LLMs fail for court forms in two predictable ways.

They are not inherently grounded in your matter data or the court’s form logic. Even when they sound confident, they can fabricate details or produce incorrect outputs. Stanford’s benchmarking work highlights hallucinations as a persistent issue in legal queries. These include high hallucination rates observed in general-purpose chatbots on legal tasks.  And the **Hallucination-Free? Assessing the Reliability of Leading AI Language Models in the Legal Domain literature frames “legal hallucinations” as a distinct reliability problem, not a minor formatting glitch. 

Even if you could “prompt your way” into decent drafts, professional responsibility does not let you outsource accountability to a chatbot. American Bar Association Standing Committee on Ethics and Professional Responsibility Formal Opinion 512 is explicit. Lawyers using generative AI must consider duties of competence, confidentiality, supervision, and candor. They must carefully review outputs to ensure they are not false, including issues like nonexistent citations and inaccurate analysis.  That’s not optional. That’s the job.

Automate Court Forms

Legacy document automation software fail differently. They can be excellent at templated document assembly, but court forms break them at the seams unless you commit to constant maintenance and deep integration.

Software like Mitratech HotDocs are built around template automation and guided “interviews” that collect variables and assemble documents. That is powerful, but it assumes you can build and maintain the template logic and that your data sources are clean and mapped.  It also means a lot of the real work shifts to template engineering, which is not free.

And Clio Draft (formerly Lawyaw) offers fillable court-form libraries and auto-population tied tightly to its ecosystem.  If your system of record is somewhere else, you often recreate the same problem: data shuffling, duplicate entry, and operational drift.

The pattern behind all these failures is the same: if the automation layer is not connected to the system where truth is captured, the user becomes the integration.

What a modern automated workflow looks like

A realistic automation workflow has to do three things well: pull reliable matter data, apply court-specific rules, and keep the loop closed inside your existing practice system.

That is exactly what MyCase Open API is designed to enable at the platform level. You connect other software to MyCase to increase data sharing and reduce manual, error-prone duplicate entry.  CaseForm then positions itself as a form-focused layer on top of that. There’s a direct MyCase integration, auto-population of form fields using existing case information, jurisdiction-specific rule application, and validation checks before filing. 

This loop is important… courts reject filings for missing info and packaging mistakes at meaningful rates, so your “automated” solution needs a rejection feedback loop, not just a PDF generator. 

Also notice what is not in the workflow. There is no copy-paste from emails into forms. Think retyping party names in multiple PDFs, and “ask the chatbot which form to use” roulette.

Case-study style examples with realistic savings

These examples are illustrative models, not promises. They are designed to be conservative on hourly cost and aggressive on accountability (assume humans still review everything). The wage baseline uses BLS median paralegal pay; attorney time is treated as scarce capacity rather than “free.” 

Example one: small family law practice dealing with frequent packets

A small family law firm prepares about 35 court-form packets per month (motions, responses, financial disclosures, service documents). Under a manual workflow, assume 90 minutes of staff time per packet for locating the right forms. Then there’s also re-entering matter data, attaching exhibits, and packaging for efiling.

At $29.33/hour, that’s about $1,540/month in direct paralegal labor for form prep alone. 

Now compare four approaches, holding review constant:

Manual:

90 minutes per packet (baseline).

Generic LLM assist:

70 minutes per packet (drafting help but still manual mapping and higher verification burden, consistent with the profession’s documented accuracy concerns). 

Legacy doc automation:

45 minutes per packet (faster assembly once templates exist, but still requires data setup and occasional manual correction). 

CaseForm-style form automation connected to MyCase:

20 minutes per packet (generation plus review and exception handling; CaseForm claims seconds-level auto-fill and built-in validation checks). 

In this model, the difference between manual and CaseForm-level workflow is roughly 41 staff hours/month. That’s about $1,200/month in direct paralegal labor. If even a fraction of that time becomes billable attorney capacity, the upside multiplies. You don’t need this to be perfect to matter because courts reject form packets at meaningful rates. Of course, rework is expensive. 

Example two: litigation practice optimizing for fewer rejected submissions

A litigation boutique files about 120 packets per month across motions and routine filings. Assume a baseline of 45 minutes of staff time per packet in a mostly-digital shop. They already e-file, but still do repetitive retyping and packaging.

That’s roughly 90 staff hours/month. At the BLS median paralegal wage, you’re at about $2,640/month in direct labor just to get filings out the door. 

Now layer in a rejection reality check… Los Angeles Superior Court’s civil rejection rate in the referenced period sits around 8–9% overall. Each rejection is at least one more cycle of staff time, attorney attention, and schedule risk.  Even if your law firm’s actual rejection rate is lower than that, a validation-driven system that catches missing fields and mismatched case data before submission is one of the few levers that reduces both time and risk simultaneously. CaseForm explicitly positions built-in validation and jurisdiction rules for that purpose. 

Speed matters, but preventing rework matters more. A faster workflow that “fails fast” into rejection is not a win. So, are you going to automate court forms at your law firm?

The post Best Way to Automate Court Forms Without Hiring More Staff appeared first on Advocate Daily.

Every motion, petition, declaration, and procedural step relies on standardized documents. The state’s court infrastructure processes millions of filings every year, and even small errors can send a case backward in the queue.

This is not a theoretical problem. According to data from the Judicial Council of California, California’s courts processed more than 5.3 million cases in fiscal year 2024–2025, serving a population of roughly 39 million people. At that scale, even minor administrative inefficiencies can ripple across the system.

The numbers show that delay is not uncommon. Statewide time-to-disposition statistics indicate that only 65 percent of general unlimited civil cases are resolved within 12 months, falling short of the judicial branch’s own target of 75 percent. Even after two years, only 86 percent of those cases have been disposed, despite a target of 100 percent.

Those delays are not caused solely by complex legal disputes. A surprising amount of friction happens earlier in the process: filing documents correctly.

And that is where a new wave of legal technology companies sees opportunity.

The hidden cost of filing mistakes

Electronic filing systems were supposed to make the legal process faster and more efficient. In many ways they have. But they have also introduced new failure points.

Court published data in California reveals how frequently filings are rejected because of relatively small errors.

For example, the Superior Court of California, County of Los Angeles has published detailed statistics on e-filing rejection rates. Between May 2023 and April 2024, the court reported average rejection rates of 17.1 percent in family law matters and 13.7 percent in probate cases.

The most common reason for rejection was documents missing information or required attachments, which accounted for roughly 35 percent of rejected family law filings.

When a filing is rejected, it does not merely require a correction. It often restarts the process. Courts may require resubmission, and clerks typically review documents on multi-day processing cycles. In practical terms, that means a filing mistake can add several days or even weeks to a case timeline.

Other courts have warned of similar issues.

The Superior Court of California, County of San Bernardino notes in its e-filing guidance that if a document is rejected and becomes untimely as a result, the filer may have to seek relief from the court to correct the error. Common rejection causes include incorrect filing codes, mismatched case information, non-compliant PDF formatting, and selecting the wrong court location.

Even when filings are accepted, attorneys remain responsible for ensuring compliance with confidentiality rules and data redaction requirements. The Superior Court of California, County of Orange explicitly warns that clerks do not review every pleading for confidential identifiers. If sensitive information is mistakenly included, the consequences can extend beyond administrative delay.

In other words, procedural precision matters.

A system built on forms

California’s courts rely heavily on standardized forms issued by the judicial branch. Thousands of different Judicial Council forms exist across civil, family, probate, criminal, and small-claims matters.

Those forms are designed to streamline the legal process, but they come with strict formatting and information requirements. Many must be completed exactly as prescribed.

To complicate matters further, forms are periodically revised. The courts publish updates and revisions several times a year, meaning attorneys must constantly ensure they are using the latest versions.

For law firms, the process often looks familiar.

Client information is gathered in one system, then manually re-entered into multiple forms. A declaration may require the same case number, party names, and addresses to be typed repeatedly across several documents. Attachments must be added, checkboxes selected, and filing codes matched to the correct document types.

When mistakes happen, the filing comes back. In a system processing millions of cases annually, these small administrative loops accumulate.

Automation enters the filing process

Against this backdrop, the legal technology company Caseway has launched CaseForm, a product designed to automate the preparation of California Judicial Council forms. The software integrates with the practice management platform MyCase and allows law firms to generate court forms directly from existing case data.

Instead of manually entering information into each document, the system pulls relevant matter details and populates the forms automatically.

CaseForm also performs validation checks before filing. These checks are intended to flag missing information, inconsistent entries, or structural errors that could trigger rejection by a court’s e-filing system. According to the company, the goal is straightforward: reduce administrative rework.

“Access to justice is not just about arguments in court,” said Alistair Vigier, CEO of Caseway. “It is about moving cases forward. In California, an enormous amount of time disappears into forms and filing corrections. CaseForm is designed to produce documents that are complete and validated before they are submitted.”

In other words, fewer rejected filings.

Why the timing matters

Legal technology has historically focused on research, analytics, and document drafting. Filing itself has received less attention. But the operational data coming from courts suggests that administrative accuracy could have a measurable impact on efficiency.

When millions of cases move through the system every year, eliminating avoidable filing errors can reduce the number of rejected submissions, repeated clerk reviews, and procedural delays.

Even modest improvements in filing accuracy could have significant downstream effects. For example, if a court with a 17 percent rejection rate could reduce that number by just a few percentage points, thousands of filings would move forward without resubmission.

That could lead to capacity improvement for the courts themselves.

The broader trend in legal technology

CaseForm reflects a broader shift in legal technology toward workflow automation.

Rather than building standalone applications, newer systems are embedding automation directly into the software lawyers already use to manage their cases.

The strategy is to reduce the number of places where lawyers and staff must manually enter the same information. For California law firms dealing with high volumes of procedural filings, the promise is appealing. Less time spent correcting administrative errors means more time focused on substantive legal work.

It also addresses that the legal profession still spends an enormous amount of time dealing with paperwork that computers are well suited to handle.

The paperwork future of law

Technology will not eliminate court backlogs on its own. Delays stem from many factors, including judicial staffing levels, case complexity, and procedural requirements.

But the data shows that paperwork inefficiencies are part of the equation. And if the legal system truly runs on forms, improving how those forms are created and submitted could become one of the most practical ways to increase efficiency.

For lawyers practicing in California’s form heavy system, that might be the most immediate impact of legal automation yet.

The post Why California Courts Are Ripe for Form Automation appeared first on Advocate Daily.

Remember the DeepSeek scare last year?

A Chinese AI model made headlines for closing the gap with US AI, and people proclaimed the end of Western dominance overnight. And then after hearing about it non-stop for two weeks, I never heard about it again.

Now Anthropic releases a GitHub repo of AI prompts, and billions in market value evaporate in a day. Thomson Reuters (the company behind Westlaw’s legal database) saw its stock plunge 18%, with RELX (parent company of LexisNexis) down 14%, and other data-heavy firms like Pearson, Sage, Wolters Kluwer, LSE Group, and Experian all taking a beating on the news.

This was because Anthropic, a large AI lab, said it can help with tasks like contract review, NDA triage (sorting non-disclosure agreements), compliance checks, and legal brief drafting. Evidently, just seeing “Anthropic” and “legal” in the same sentence was enough to trigger a sell-off. The irony is rich. Investors panicked over capabilities that any law student or junior attorney would find rudimentary.

A Reality Check on the “Breakthrough”

Here is a dose of realism. Anthropic’s so-called breakthrough was essentially a collection of open-source prompt templates. There was no new model and no proprietary data. The core method for contract review was straightforward. It reads the whole contract, check key clauses, and compare against a playbook.

Shocking, I know.

As a side note, this was likely done to get into the AI contract-review space that startups like Spellbook and Harvey occupy. It’s not what Caseway does. We don’t do contract review, NDA triage, compliance checks, and legal brief drafting.

The market’s knee-jerk reaction says more about our collective short-term memory than about the technology itself.

Validation, Not a Surprise

Why We Stayed Quiet

At Caseway, we have been notably silent during these hype cycles. This was not due to a lack of awareness. It was because we were busy building propriatary data. When everyone else scrambled to respond to the latest GPT release or declared an AI bubble, our team doubled down on research and product development.

We chose our direction long before Anthropic decided to open-source their legal plugin. Their move validates our early bets. It confirms what we believed from the start. And that is to be ready to transform legal work, form automation, and enterprise data workflows using the companies’ own data. Their entry accelerates momentum in areas we were already building toward.

We did not publish hot takes on DeepSeek or rush out press releases about Anthropic’s software because we do not react impulsively. We prepare deliberately. Unlike others, we research and understand what Anthropic is actually offering.

Building the Hard Things First

Academic Partnerships and Long-Term Research

Over the last year, Caseway quietly forged strategic partnerships with leading universities such as UBC and SFU to address foundational problems in legal artificial intelligence.

These are multi-year, government-funded research projects designed to solve issues that hype alone cannot. With UBC’s NLP group, we are working on reducing AI hallucinations, including fake legal citations and misinformation, in legal research systems. This work is backed by a national research grant.

At SFU, we partnered on a project to index 100 million court decisions and make them machine-readable. The goal is to evaluate whether improved AI access to case law can materially improve outcomes for people without legal representation. The work we do will be sourced by ChatGPT and Anthropic.

These efforts signal to serious observers that Caseway has been methodically building a defensible position rooted in accuracy, compliance, and real-world deployment. Other companies that built ChatGPT and Anthropic wrappers are going to get hit hard.

Anthropic Confirms the Thesis

Anthropic’s move effectively proves our point. Legal departments want AI. General AI providers see the opportunity. The difference is timing and depth. We have been building these solutions quietly for over a year.

While others live-tweeted their pivot to AI, we integrated our technology into institutions that cannot afford unreliable systems. We stayed focused. That focus is now paying off.

Why Domain-Specific AI Still Wins

Specialization Is the Advantage

One lesson repeatedly lost in the hype is that domain expertise matters. A one-size-fits-all model struggles with the nuance of specialized fields. Even a big-tech AI entrant in legal would hit the same wall… Without specialized training and data, they can’t capture the nuance. This is exactly the edge Caseway has built for the legal domain.

This has been our position from day one, and it is gratifying to see the market catching up.

Anthropic’s entry into legal AI reinforces the path we took with our product suite: Casey, CaseForm, and Synthium. These are purpose-built systems, not generic chat interfaces.

Below is why that distinction matters.

Casey: Legal Research Built on Ground Truth

Casey is trained exclusively on verified legal sources (court decisions, statutes) and nothing else. It won’t tell you the latest sports scores or generate a recipe, and that’s by design. By limiting its knowledge base to what judges actually wrote, Casey avoids the pitfalls that plague general models. This includes hallucinations and inaccuracies on legal questions.

We all heard about the lawyer who cited fake cases from ChatGPT in a legal brief, embarrassing himself because the AI invented cases that don’t exist. Casey was built so that can’t happen: “If a judge didn’t write it, our AI doesn’t know it,” as I like to say. 

Ask Casey something outside real case law, and it will politely admit it has no idea, rather than fabricate an answer. Every response is backed by pinpoint citations to the actual source. This means attorneys using Casey can instantly verify the AI’s answer against the original judgment or statute. This is a must in law, where trust is everything.

Casey was designed with privacy from the ground up. We retain zero user data, and offer on-premises deployments for firms that need full control. Compare that to a general AI software where your queries might be logging who-knows-what on some server. It’s no wonder cautious law firms and government agencies prefer a “privacy-first” solution like Casey over a generic chatbot. When Anthropic’s general model jumps into legal, it confirms there’s a big market here… But it also highlights what we’ve built: an AI that actually understands legal nuance and won’t recklessly spit out nonsense.

CaseForm: Automation That Survives the Real World

If you’ve ever dealt with legal or government forms, you know they are a special kind of pain. They come with strict formats, jurisdiction-specific rules, and hidden pitfalls that a generic AI wouldn’t catch. CaseForm tackles this head-on. 

It doesn’t just “fill PDFs” like some basic script; it intelligently reads your input documents, understands the context and rules (e.g. court form requirements or regional regulations), and produces fully-completed forms that won’t get rejected for a clerical error.

Under the hood, we baked in legal domain knowledge. These are things like jurisdiction-specific validation rules and compliance checks, that a one-size AI wouldn’t know. When Anthropic’s demo shows an AI can draft a template response or flag a clause, we applaud.

CaseForm goes further by ensuring the generated document is actually compliant with the relevant legal standards (say, the correct format for a federal filing, or the right waiver language for a provincial court).

AI without a massive IT project

This kind of built-in expertise is not something you get by dumping the internet into a model. It comes from partnering with legal professionals and understanding the workflows intimately. Another key difference is how CaseForm fits into real organizations. 

We made it easy to integrate. It doesn’t matter if it’s a law firm wants to plug it into their case management software or a bank wants it in their compliance pipeline. The idea is that a legal department can get quick wins from AI without a massive IT project.

Generic software might give you a flashy demo of filling one form; CaseForm is about seamlessly automating thousands of forms in your live environment, with enterprise support on standby. In the wake of Anthropic’s launch, we expect document automation demand to surge, and we’re ready for it. CaseForm has been riding this wave for a while, and our focus now is keeping that edge by folding in the latest AI advances while emphasizing the domain-specific smarts that generic software lacks.

Synthium: Bringing AI Inside the Firewall

The Anthropic news wasn’t just about legal research; they also open-sourced tools for things like data analysis and support. It underscores a broader trend that companies want to bring AI into their own data stacks, not just use third-party APIs. Synthium is our answer to that. It’s an enterprise data hub that marries AI with your internal data. 

Here’s why open-sourcing actually plays into Synthium’s hands. Anthropic made their legal AI plugin open-source and customizable, which is great. But let’s be real… How many companies have the in-house AI engineers to safely deploy and tailor those open software across their proprietary databases?

Very few. It’s one thing to read a GitHub prompt; it’s another to wire up an AI that combs through millions of your documents, respects all your privacy and compliance requirements, and plugs into SharePoint, Salesforce, or whatever systems you use. Synthium provides that “glue” out-of-the-box. 

We offer secure, governed AI deployments where the data stays within your firewall. No sending your sensitive info off to some external API and crossing fingers it won’t leak. With Synthium, a bank or hospital can harness AI on their datasets with full auditability. We’ve built in connectors to the usual enterprise systems. This includes databases, cloud storage, ERPs, so you don’t have to rip out your existing workflows.

Configure custom pipelines

Crucially, we also bring our team to the table: integration specialists who help configure custom pipelines and make sure the AI implementation actually delivers value. This is the unsexy part of AI adoption that outsiders often ignore. Anthropic’s open-source release lowers the barrier to experimenting, sure. 

But when a Fortune 500 firm actually tries to roll it out, they’ll hit the barrier of complexity and risk. That’s where Synthium shines… It’s a ready-made platform combining the latest open AI innovations with enterprise-grade reliability, support, and compliance features.

The scare Anthropic gave to big data incumbents could drive more businesses to seek out solutions like Synthium. Companies now realize AI is a competitive necessity, and many will prefer to invest in their own data strategy (where they keep control) rather than rely purely on external software. 

Our bet is that enterprises will want the best of both worlds. This includes the power of cutting-edge AI and the safety nets of a proven, private system. Synthium was built for exactly that scenario, and the latest news just accelerates the timeline.

Open Source as an Accelerator

Anthropic open-sourcing its legal AI software does not commoditize what we do. It reinforces it. The release makes clear that value lives in implementation, integration, and trust. 

Anthropic shared a blueprint. We can incorporate useful components faster than closed competitors, while continuing to differentiate on data curation, experience, compliance, and support. Those elements are far harder to replicate, forming a durable moat around Caseway’s business that general-purpose AI initiatives can’t easily cross. Those elements are far harder to replicate.

Organizations experimenting internally will discover that turning raw prompts into reliable systems is non-trivial. When that realization hits, they will look for experienced partners. We have been doing this work for years.

The industry needs to stop treating every open-source release as an existential threat. If your product is a thin wrapper, concern is warranted. If your strength lies in domain depth and delivery, open innovation expands the opportunity.

Hype Fades, Fundamentals Remain

The market reaction to Anthropic’s launch underscores that AI’s impact is real, hundreds of billions of dollars real. The mistake is assuming each announcement reshuffles the entire hierarchy overnight.

Incumbents relying on legacy advantages are vulnerable. Empty AI startups with no differentiation are also vulnerable. Teams that quietly built defensible systems grounded in real use cases are in a strong position. So to investors and founders: pause the drama. Focus on fundamentals. Who controls the data and has earned trust? Who can deal with regulation without slowing innovation?

Those questions determine long-term winners.

Anthropic’s Legal AI

Caseway has been answering them consistently, away from the hype cycle. We welcome competition because it validates the market. Caseway embraces open source because it accelerates progress. We partner because durable advantages are rarely built alone. Disruption is about separating substance from noise. The companies with real foundations endure. The noise fades.

Caseway’s path, measured, research-backed, and customer-driven, may not dominate headlines weekly (although we are in the news when it matters.) But as this episode shows, when the dust settles, quiet execution tends to win.

Bet on execution. Every time.

Caseway is proud to partner with Northeastern University on applied research that actually matters.

This capstone project, Zero Trust in Practice, dives into the hard, unglamorous work of securing real infrastructure. PostgreSQL, Elasticsearch, and object storage. Real systems, real tradeoffs, real latency costs. Exactly the kind of research we care about as we continue hardening the security and privacy foundations of Caseway’s products.

The team explored how Zero Trust architectures can be implemented across heterogeneous data systems using centralized policy enforcement, evaluated performance impacts, and surfaced the uncomfortable truth most vendors avoid.

Strong security is never free, but done right, it is predictable, auditable, and worth the cost. That thinking directly aligns with how we build at Caseway.

I am genuinely excited for the current AI hype bubble to end.

When the noise dies down, when the market gets tired of ChatGPT wrappers duct-taped onto sensitive workflows, companies that are deeply technical, security-first, and infrastructure-aware will finally be heard. Caseway is built for that next phase. Quietly. Deliberately. With real engineering under the hood.

Huge thanks to the Northeastern students who worked on this project. You tackled a complex problem with rigor, clarity, and professionalism. This is the kind of work that moves the industry forward, even if it does not fit neatly into a demo reel.

More of this. Less hype.

The post Anthropic’s Legal AI Launch Confirms Caseway’s Strategy appeared first on Advocate Daily.

Courts, lawyers, and the public are increasingly relying on AI tools to answer legal questions, yet many of these systems are trained on scraped internet content rather than official court decisions. The result has been hallucinated citations, fabricated precedents, and growing concern about whether AI can be trusted in high-stakes legal settings.

In response, the University of British Columbia and Vancouver-based startup Caseway have launched a two-year research collaboration to build a legal AI system grounded exclusively in real court decisions. The project aims to dramatically reduce hallucinations and improve the reliability of legal research AI in Canada.

As General-Purpose AI Pulls Back from Legal Use

The timing of this initiative is significant. In November 2025, OpenAI updated ChatGPT’s usage policies to prohibit the AI from providing legal advice, a change that has stirred debate across the technology industry. This retrenchment by the world’s most popular AI chatbot has raised concerns about a new gap in legal help for the public. Many individuals, unable to afford a lawyer who might charge $400/hour, had begun turning to AI for guidance on legal questions.

If general AI models step back from legal assistance, who will fill that void? The UBC–Caseway team believes domain-specific AI is the answer. Building a custom LLM-based legal software is viewed as necessary to overcome the limitations of general-purpose models like ChatGPT, including their tendency toward errors and misinformation. By focusing narrowly on legal knowledge and reasoning, the new model aims to provide dependable answers that ordinary people can use with greater confidence.

Caseway’s database already indexes millions of court decisions

“Access to justice is a huge problem, here in Canada, roughly 70% of people in the justice system lack legal representation,” said Caseway’s Alistair Vigier. “With ChatGPT pulling back, that’s one less tool people might have experimented with for help. We see our legal AI as a way to level the playing field. Whether you’re a lawyer at a law firm or someone preparing for court on your own, this technology will offer reliable legal research at your fingertips, so you’re not left in the dark.” 

The need is indeed pressing: a majority of those navigating civil and family courts on their own often struggle to interpret laws and find relevant precedents. 

Vigier says that if everyday Canadians could ask an AI trained on official court decisions and get answers without the usual AI fabrications, it could empower them to understand their rights and options better. He also points out that misinterpreting the law or missing a key case can have life-changing consequences.

“Getting the right information quickly can make all the difference. We’re not building this to replace lawyers, but to make legal knowledge far more accessible to those who can’t easily get professional help.

The problem

Every day, thousands of Canadians appear in court without a lawyer. Many turn to online searches for help, and find misinformation instead. The Caseway-UBC partnership project aims to change that.

The project’s goal is to dramatically reduce AI “hallucinations,” instances where generative AI produces false or made-up information, and improve the accuracy of legal research software like Caseway. By combining UBC’s cutting-edge natural language processing research with Caseway’s real-world legal data and expertise, the team plans to create one of the first dedicated legal artificial intelligence models in the world. This home-grown Canadian initiative comes at a pivotal moment, as mainstream AI systems grapple with reliability issues and new restrictions on giving legal advice.

Across the world, governments and researchers are wrestling with how to make AI more accountable. In fields like law, where precision can determine someone’s freedom, financial security, or access to justice, even one false citation can have real-world consequences.

Two-Year NSERC and Mitacs Project Tackles AI Hallucinations in Law

The partnership brings together Dr. Vered Shwartz, an Assistant Professor of Computer Science at UBC and Canada CIFAR AI Chair, and Alistair Vigier, the CEO and co-founder of Caseway. Backed by research grants from NSERC and Mitacs, their joint project will span two years and support a postdoc and a graduate student  to develop an accurate and reliable LLM-based legal research software. 

Uniquely, the tool will be built on open-source AI foundations, allowing the team to customize it for legal applications while ensuring transparency and potential for wider community use.

“This partnership is a game-changer for legal technology,” said Alistair Vigier, CEO of Caseway. “Combining Caseway’s practical innovation with UBC’s world-class AI research, we’re poised to create something truly transformative. We want to make sure anyone, from lawyers to self-represented people, can get reliable legal answers when they need them.”

Natural language processing and responsible AI

Dr. Shwartz’s lab specializes in natural language processing and responsible AI applications in sensitive domains like law. One of the central challenges the team will tackle is the “hallucination” problem in current AI. General-purpose models like ChatGPT have been prone to confidently output incorrect legal citations or non-existent precedents, which can mislead users. These mistakes by AI pose obvious risks in law, where accuracy and truth are paramount. 

“LLMs are already being used in fields such as law, but they still suffer from limitations like overconfidence and the tendency to hallucinate facts,” Dr. Vered Shwartz noted, citing her concern that rushing AI into legal use without fixes could be harmful. 

“People are already using LLMs to get answers for legal questions. Our goal is to deliver these answers in the context of Canadian law, with source attribution and  reducing  hallucinations. By  focusing on tasks such as improving the granularity of the retrieval of case laws, combining LLMs with symbolic reasoning, and rigorously testing the tool, we aim to create an AI that professionals and the public can trust to provide accurate legal information,” said Shwartz. 

Towards Affordable and Inclusive Legal Research

Both UBC and Caseway view this collaboration as an opportunity to advance access to justice, particularly for marginalized communities. Caseway has also been in discussions with groups such as the Indigenous Bar Association on how AI software could help address unique legal information needs in Indigenous communities.

Cultural and linguistic nuances, which Dr. Shwartz’s research also emphasizes, will be considered in training the model, so that it can serve diverse user groups responsibly. “We want to ensure this technology benefits all Canadians, including those who have historically faced barriers in the legal system,” Vigier said. 

In line with that mission, Caseway is planning a shift to a freemium model for its services in the future. While today a subscription is required, the company is actively seeking grants and public funding to offer core legal research features for free to the public. 

Caseway notably avoids traditional venture capital and instead leverages angel investors and Canadian government grants to maintain an access-focused approach. The development of a free, AI  legal research platform is already underway, which would allow anyone to pose legal questions and receive answers backed by actual case law at no cost. “Our vision is to provide the average person the same research power that lawyers have, without the hefty price tag,” said Vigier, noting that sustainable grant support will enable the company to keep the lights on while offering a basic version of the software to all.

A Breakthrough for Legal Tech and Academia

This UBC–Caseway partnership highlights how academia and industry can join forces to solve real-world problems. It also positions Canada at the forefront of legal AI innovation.

By building a specialized legal research tool that prioritizes factual accuracy and accountability, the team hopes to set a new standard for AI in the legal field. Media watchers have noted that the project’s objectives align with growing calls for responsible AI development, creating systems that are not only powerful, but also transparent and safe for public use. 

If successful, the legal AI developed here could become a blueprint for similar efforts in other countries and domains. For now, the collaboration is just beginning, with research work kicking off this fall at UBC’s Vancouver campus and Caseway’s data engineers working in tandem. All eyes in the legal tech community will be on this initiative over the next two years. 

How legal research is conducted

A successful outcome could fundamentally change how legal research is conducted, slash the time and cost required to find answers, and open up the justice system to millions more. As major AI companies retreat from high-stakes domains like law and medicine, Canadian innovators are stepping up to build domain-specific systems that prioritize accuracy over scale.

If successful, this initiative will not only transform legal research but also demonstrate that Canada can lead the world in responsible AI. “We’re building a model that reflects Canadian values of fairness, transparency, and inclusion,” said Alistair Vigier. “The law should be understandable to everyone, and this technology brings that goal within reach.”

The post When Legal AI Hallucinates, Justice Suffers appeared first on Advocate Daily.