Legal AI in 2025: The Year the Money Got Serious
Six billion dollars in funding, a string of mega-partnerships, and 660 documented hallucination cases. A month-by-month account of the year that reshaped legal technology.
I have been watching technology reshape professional services for three decades. I have never seen a year like 2025.
Not because the technology was unprecedented, though it was impressive. Not because the money was staggering, though six billion dollars is a number that demands attention. But because 2025 was the year when the legal profession collectively crossed a line. AI stopped being something to debate and became something to deploy. Whether organisations were ready for that crossing is another matter entirely.
Here is what happened, month by month.
Q1: The Stage Is Set
January opened with the new Trump administration signalling a lighter regulatory touch on AI. For European observers like me, this was significant not because of what America did, but because of the contrast it created. The EU was tightening. America was loosening. Organisations operating across both jurisdictions suddenly had two very different compliance frameworks to navigate.
Meanwhile, the hallucination count stood at roughly 120 documented cases. That number would look almost quaint by December.
February brought the ABA Commission's working group recommendations, establishing what they described as "immediate attorney obligations" for AI use. This was the American legal establishment drawing a line: you can use AI, but you must understand what you are doing. For anyone following the Article 4 developments in Europe, the parallels were unmistakable. Different jurisdictions, same underlying concern.
March delivered the first major funding event of the year. Harvey closed a three hundred million dollar Series E at a five billion dollar valuation. I remember reading that number and doing the arithmetic. Three years old. Five billion. In legal technology. I have watched companies in this space grow for decades, and the velocity was genuinely without precedent.
Q2: The Mega-Deals
April was when the investor confidence became impossible to dismiss. Eve, focused on plaintiff-side litigation, raised one hundred and three million at a billion-dollar valuation. The LegalTech Fund closed its second fund at one hundred and ten million, nearly four times its inaugural fund from 2021.
What struck me about the Eve round was what it represented. Plaintiff-side litigation is adversarial, high-stakes, and traditionally resistant to technology adoption. The fact that serious investors were backing AI in this space told me the adoption wave had moved well beyond contract review and due diligence.
May brought something genuinely new: Garfield received SRA authorisation as the first AI-only law firm in the United Kingdom. I do not think most people appreciated the significance. A regulatory body that traces its oversight of legal practice back centuries formally approved a firm where artificial intelligence, not human lawyers, performs the substantive work. Whether Garfield succeeds commercially matters less than the precedent.
June was arguably the most consequential month of the year. Harvey and LexisNexis announced a strategic partnership that Richard Tromans, one of the sharpest analysts in legal technology, called "possibly the most important legal tech move in a decade." At the same time, Thomson Reuters unveiled agentic CoCounsel, describing it as a "fundamental shift from AI assistants to intelligent agentic systems."
The language mattered. We had moved from "tools" to "agents" to "systems." Each word carried more autonomy, more ambition, and more risk.
Q3: The Agents Arrive
July was a month of quiet regulatory compliance, as multiple state AI regulations approached implementation deadlines. The headline events were elsewhere, but the unglamorous work of mapping requirements to systems was happening in legal departments across the country.
August was ILTACON, and the theme was unmistakable: agentic AI. Litera unveiled Lito. NetDocuments launched AI-powered document profiling. Thomson Reuters released CoCounsel Legal. Every major vendor had an "agent" story.
I attended presentations and watched demos, and I kept thinking about the same question. An agent, by definition, acts with some degree of autonomy. In legal work, autonomy without verification is negligence. The technology was advancing faster than the governance frameworks needed to contain it.
August also brought the EU AI Act's first obligations into effect for general-purpose AI models. The regulatory clock was ticking.
September gave us Clio's extraordinary fundraise: eight hundred and fifty million in combined venture and debt financing, following nine hundred million in 2024. And Cleary Gottlieb, one of the most prestigious firms in the world, acquired UK startup Springbok AI. When Cleary Gottlieb makes a buy-versus-build decision, the market pays attention.
Q4: Consolidation and Reckoning
October opened with what was described as the largest acquisition in legal technology history. Clio announced a one billion dollar acquisition of vLex. The consolidation I had been expecting was arriving faster than I predicted.
And then Harvey reached an eight billion dollar valuation. Doubled from five billion in six months. In a legal technology company that was three years old.
November delivered both ambition and caution. Norm Ai launched Norm Law LLP, an AI-native law firm targeting institutional clients. Fascinating model. At the same time, reports emerged that Robin AI was encountering funding difficulties, and that dozens of legal tech companies that raised capital between 2020 and 2023 had not raised again. The market was bifurcating: massive capital for the leaders, silence for the rest.
December provided the year's bookends. The ABA Task Force released its Year 2 Report, concluding that AI had moved "from experiment to infrastructure." The rate of documented hallucination cases had accelerated to four or five new incidents per day, bringing the total past 660. And President Trump signed an executive order creating an AI Litigation Task Force to challenge state AI laws.
The tension could not have been starker. The profession's most authoritative body declared AI to be infrastructure. The evidence of AI's unreliability grew faster than ever. And the regulatory environment remained fractured between innovation-friendly federal policy and increasingly prescriptive state requirements.
The Numbers That Matter
Total legal tech funding in 2025: $5.99 billion across fourteen rounds exceeding one hundred million dollars. Harvey alone raised seven hundred and sixty million across three rounds in a single year.
By December, 79% of law firms had integrated AI tools. Fifty-five percent of law schools offered AI-focused courses. Eighty-three percent provided hands-on AI experiences. Ninety-one percent of state bars were developing AI guidance.
And 660 documented cases of AI-generated fabrications in legal filings.
What 2025 Actually Taught Us
The numbers tell one story. The year, lived through, tells another. What I took away from 2025 was not that legal AI arrived. It was that legal AI arrived without the profession having resolved the fundamental questions about how to use it responsibly.
The Stanford research showing 17 to 33 percent error rates for major AI platforms was published the same year that firms deployed those platforms to hundreds of lawyers. The hallucination count quintupled in twelve months while adoption accelerated. The money flowed in one direction while the evidence pointed in several.
This is not a contradiction. It is how technology transitions actually work. Adoption and readiness are different timelines. The organisations that succeed are the ones that close the gap between the two, not by slowing adoption but by accelerating readiness.
That gap, between what AI can do and what organisations are prepared to manage, is the defining challenge of 2026.
Alex Blumentals is the founder of Twin Ladder. He has watched three decades of technology transitions and finds 2025 to be the most consequential year for professional services since the arrival of the commercial internet.