Legal AI in Q1 2026: Competition shifted from the model to legal content, playbooks and matter systems
Shares of Thomson Reuters, RELX and Wolters Kluwer fell 10.5% to 18% on Feb. 3 after Anthropic released a free legal plug-in, and by Feb. 25 LexisNexis and Thomson Reuters had both announced plans to build on Anthropic's technology. For firms and legal departments, the quarter brought broad AI use with little change in pricing or policy, and courts that began enforcing the duty to verify what is filed.
Editorial snapshot: 30 April 2026
The central finding
The value in legal AI moved toward whoever controls curated law, client playbooks and the systems where matters live, while responsibility for checking AI output stayed with the lawyer, and courts began to enforce it.
What changed in legal AI in the first quarter of 2026
Four developments define the quarter. A model developer packaged first-pass contract work into a free agent, private capital flowed to two AI-native platforms, use spread faster than firms changed how they charge, and courts turned verification into an enforceable duty.
Capability
A general agent took on in-house contract work
Anthropic's open-source legal plug-in, released Jan. 30, reviews contracts and triages nondisclosure agreements against a team's own playbook through connectors to email, storage and e-signature, with no case-law connector. On Feb. 25, LexisNexis said it would integrate the plug-in into Protégé, and Thomson Reuters said its next CoCounsel Legal is built on Anthropic's Claude.
Capital
Private capital concentrated in two platforms
Legora raised $550 million at a $5.55 billion valuation on March 10, and Harvey raised $200 million at $11 billion on March 25; Law360's tracker credits the pair with nearly half of first-quarter legal-tech funding. The median round in Legal Complex's count of 103 deals was $1 million, and the largest item was Relativity's $720 million term loan, used partly to refinance existing debt.
Adoption and pricing
AI use outran changes in revenue and billing
In a Clio survey released March 9, 71% to 78% of solo and small practices and 86% of midsize firms use AI, but only 31% to 33% of the smaller firms say it has improved revenue, against 59% at enterprise firms. About 90% of legal dollars are still billed hourly, according to Thomson Reuters and Georgetown Law.
Accountability
Courts made verification enforceable
A District of Kansas standing order of Jan. 28 lets the court strike unverified AI-assisted filings and require a sworn account of AI use. On March 13, the 6th U.S. Circuit Court of Appeals fined two lawyers $15,000 each for more than two dozen fake or misrepresented citations, without deciding whether AI produced them.
Operating question
Which of your legal work depends on context you control, such as playbooks, precedent and matter files, and which on content you license, and who verifies each AI-assisted output before it reaches a client or a court?
A general-purpose agent reached in-house contract work
The quarter's frontier releases made million-token context and computer use standard: Claude Opus 4.6 on Feb. 5, Claude Sonnet 4.6 on Feb. 17, Gemini 3.1 Pro in preview on Feb. 19 and GPT-5.4, OpenAI's first general-purpose model with native computer use, on March 5. In legal work, the structural change was packaging. A general agent that reads local files, follows a team's written positions and reaches document stores through standard connectors now covers first-pass contract work that several products sell, and within four weeks both large content incumbents had moved to absorb it.
In-house contracts · Open source, Jan. 30
What the legal plug-in does
Built for Anthropic's Cowork desktop agent and usable in Claude Code, it reviews contracts against a playbook, sorts NDAs into green, yellow or red, checks vendor agreements, and drafts briefings and templated responses. Its skills also cover privacy work under GDPR and CCPA. The team's positions on liability caps, indemnities, governing law and NDA defaults sit in a local Markdown file the team edits.
Connectors
Connected to office systems, not to legal content
The plug-in reaches Slack, Box, Egnyte, Microsoft 365, Atlassian, Gmail, Google Calendar and DocuSign through Model Context Protocol connectors, a standard way for an agent to use other software, with optional links to Ironclad and Agiloft. It lists no legal-research or case-law connector, carries an Apache 2.0 license and states that licensed attorneys should review its output. Anthropic added a jurisdiction disclaimer on March 18.
Incumbent response · Feb. 25
LexisNexis and Thomson Reuters absorbed the agent layer
LexisNexis said it would integrate the plug-in into Protégé so outputs draw on Lexis content, for example checking a data protection agreement against its repository. Thomson Reuters said it is an Anthropic customer building next-generation CoCounsel Legal on Claude and Anthropic's Agent SDK, adding Westlaw and Practical Law content, verification and workflow controls. Anthropic announced LegalZoom and Harvey connectors the same day.
Unit cost
Model prices are public; legal AI prices are not
Legal AI vendors published no list prices for their products. Thomson Reuters says it does not sell seats, prices to value and expects a slightly higher pricing yield in 2026.
$2.50/$15GPT-5.4, per million input/output tokens
$5/$25Claude Opus 4.6, per million input/output tokens
$0.25/$1.50Gemini 3.1 Flash-Lite, per million input/output tokens
Cost and consequence
Because every vendor and many legal departments can buy the same models, a legal AI product's price now reflects the context it adds: licensed law, citators, a client's playbook and matter data. The useful comparison is the time and cost of a completed contract review or filing, including the lawyer's check, and buyers will have to measure that themselves.
Legal-tech capital concentrated in Legora and Harvey while the typical round stayed small
Legal tech's money is small beside the frontier: Crunchbase counted $300 billion of global venture funding in the quarter, 80% of it for AI, and Anthropic alone raised $30 billion on Feb. 12. Within legal tech, totals depend on the tracker. Legal Complex counted $2.34 billion across 103 deals, including debt; Legal Technology Hub counted $1.42 billion in 35 rounds, roughly a third of its $4.28 billion for 2025. Under either total, money went two ways: very large rounds for two AI-native platforms and a long tail of small seed deals.
Disclosed financing
Largest disclosed legal-tech funding rounds, Q1 2026
Disclosed amounts in USD millions, Jan. 1–March 31, 2026. Relativity's $720 million term loan is excluded because it is debt, used partly to refinance an existing private credit facility.
Legora
Series D — $550M
Led by Accel at a $5.55 billion valuation; March 10
Harvey
Funding round — $200M
Co-led by GIC and Sequoia at an $11 billion valuation; March 25
Fieldguide
Funding round — $75M
Governance, risk and compliance
Factify
Seed — $73M
Orbital
Series B — $60M
Real estate law; Jan. 26
Lawhive
Funding round — $60M
Legal services
Crosby
Funding round — $60M
Contracts; closed March 31
Ivo
Funding round — $55M
Summize
Funding round — $50M
Steno
Series C — $49M
Court reporting
Company announcements; Legal Complex's first-quarter tally via Artificial Lawyer; LawSites. Instruments as reported; "Funding round" where none was stated.
Nearly half
Share of first-quarter legal-tech funding raised by Legora and Harvey, by Law360's count
$1M
Median round across Legal Complex's 103 first-quarter deals
AI-native platforms
Valuations stepped up within months
Harvey's valuation rose from $8 billion in December 2025 to $11 billion in March, and it has raised more than $1 billion in total. Legora's $5.55 billion valuation compares with $1.8 billion at its $150 million Series C in October 2025. Neither company disclosed revenue.
Early-stage market
Seed deals outnumbered growth rounds
Legal Complex counted 46 seed deals and 44 growth deals, the first quarter since early 2024 in which seed led. Raymond Blyd of Legal Complex said 979 companies that raised seed rounds from 2023 to 2025 had not raised since.
Listed incumbents
Shares fell while legal revenue grew
On Feb. 3, shares of Thomson Reuters fell 14% to 18%, RELX 14% and Wolters Kluwer 10.5% to 13%, depending on listing and on intraday or closing prices; the S&P 500 software and services index lost about $830 billion in market value from Jan. 28 to Feb. 4. Thomson Reuters' Legal Professionals unit grew 9% organically in the fourth quarter of 2025, and RELX's Legal division grew underlying revenue 9% and profit 12% in 2025.
Exits and data
An IPO filing and a lawsuit over case-law access
By March 23, Relativity had confidentially submitted a draft IPO registration; it would be the first legal-tech IPO since 2021. In January, Alexi counterclaimed that Clio's Fastcase and vLex acquisitions violate the Clayton Act, saying its case-law feed stopped in December and it cut two-thirds of its staff; Clio calls the claims baseless. BriefCatch bought WordRake, Filevine bought Pincites and Haveli took a majority stake in Sirion, all at undisclosed prices.
What it signals
Investors are paying for platforms that legal teams build their own agents on, while public markets questioned incumbents that were still growing. For buyers, the pattern is a supplier-risk question: below the top two, rounds are small and many recent seed companies have not raised again, and one vendor's access to case law is now in litigation. Data export, continuity and content-access terms belong in AI contracts.
Legal AI now reaches buyers through four routes
Product strategy split by who owns what. Content incumbents tied workflow platforms to their own citators and content. AI-native platforms sold configurable agents. A model developer released a free workflow, then became supplier and channel to the incumbents. Practice-management vendors became the route into small firms. In personal injury and consumer claims, agents now act by phone and with insurers.
Content incumbents
Workflow platforms tied to proprietary content
Lexis+ with Protégé replaced Lexis+ AI on Feb. 24 with more than 300 workflows, a no-code builder, Shepard's-based citation checking, models from Anthropic, Google and OpenAI, and a 200-billion-document graph. Thomson Reuters described Westlaw Advantage as an agentic research product, promised an agentic CoCounsel Legal by midyear and said it is developing its own language model. It spends more than $200 million a year on productized AI.
AI-native platforms
Configurable agents, reported by customers rather than revenue
Harvey reported more than 25,000 custom agents, including long-running ones for work such as fund formation, used by more than 100,000 lawyers at 1,300 organizations, among them most of the Am Law 100 and more than 500 in-house teams. Legora reported about 800 customers in more than 50 markets and growth from 40 to 400 staff in a year.
Small firms
Practice-management systems became the channel
In late March, Smokeball and Thomson Reuters announced a CoCounsel integration, with a first phase planned for late spring or early summer, and Centerbase launched a native integration of NetDocuments' ndMAX, putting AI inside the systems where small and midsize firms keep matters and billing. RELX said large firms use more than 100 workflow software vendors, that it earns virtually nothing from workflow software and that it has about 25 such partnerships in Legal.
Personal injury and consumer claims
Agents that call, request records and negotiate
EvenUp's voice and text agents, launched Jan. 28 with early deployments, open claims, confirm coverage, check treatment, request records and verify medical balances across many cases at once. Mighty launched a free consumer platform the same day that values motor claims, assembles a settlement package and negotiates with the insurer; network attorneys pay it a lead-generation fee.
AI-native law firms
Firms that sell AI-first work per matter
General Legal, founded Jan. 7, has AI do first review and routing and charges about $500 per contract review and $250 per NDA; it reported 110 to 115 clients and about $1 million in annualized revenue by March 31. In the U.K., LawFairy sells immigration-eligibility reports for £149 from a rules engine that uses no large language model.
Status of selected legal AI releases in the quarter
Dates and status as announced; announced agentic, beta and phased releases had not shipped by March 31.
| Release | Date (2026) | Status in the quarter | What it does |
|---|---|---|---|
| Relativity aiR for Case Strategy | Date (2026) Jan. 12 | Status in the quarter Launched | What it does Extracts and organizes key case facts |
| LexisNexis Protégé AI Workflows | Date (2026) Jan. 21 | Status in the quarter Commercial preview | What it does Prebuilt workflows and a custom builder |
| EvenUp communication agents | Date (2026) Jan. 28 | Status in the quarter Early deployments | What it does Voice and text agents for claim tasks |
| Mighty | Date (2026) Jan. 28 | Status in the quarter Launched to consumers | What it does Free motor-claims resolution |
| Anthropic legal plug-in for Cowork | Date (2026) Jan. 29–30 | Status in the quarter Open source; Cowork in research preview | What it does Playbook-based contract review, NDA triage |
| DISCO agentic tool | Date (2026) Feb. 9 | Status in the quarter Announced; due later in 2026 | What it does Agentic review for large discovery matters |
| Harvey and Intapp Walls | Date (2026) Feb. 23 | Status in the quarter Announced | What it does Ethical-wall policies applied to Harvey access |
| Lexis+ with Protégé | Date (2026) Feb. 24 | Status in the quarter Generally available in U.S.; global through 2026 | What it does 300+ workflows; agentic workflows to follow |
| Next-generation CoCounsel Legal | Date (2026) Feb. 24 | Status in the quarter Beta due soon; agentic version promised by midyear | What it does Built on Claude and the Agent SDK |
| American Arbitration Association Resolution Simulator | Date (2026) March 4 | Status in the quarter Launched | What it does Nonbinding simulated outcomes, documents-only disputes |
| CoCounsel in Smokeball; ndMAX in Centerbase | Date (2026) March 24–25 | Status in the quarter Integrations announced | What it does AI inside practice management |
Implementation
Each route puts the engineering work in a different place. Platforms need connections to a firm's document, matter and email systems; the plug-in needs a team to write its own playbook and set file permissions; practice-management integrations bundle both for smaller firms. Whichever route a firm takes, ethical walls and audit logs now have to cover what an agent can open, as Harvey and Intapp's integration does for Harvey.
AI use is broad across firm sizes; revenue gains and formal policy concentrate at larger firms
Reported use depends on what a survey counts. In 8am's survey, 69% of individual legal professionals used general-purpose AI, but only 46% of their firms provided such tools and 34% provided legal-specific ones. Most fieldwork predates January. The sharper differences lie in what firms did after adopting: formal policy and reported revenue gains both rise with firm size.
Reported AI use, by survey and population
Each publisher defines its own population and measure, so the rows are not a trend or a single scale. Clio and 8am survey the firms they sell to.
| Population measured | Reported use | Who was surveyed | Fieldwork |
|---|---|---|---|
| Individuals using general-purpose AI (8am) | Reported use 69% 31% a year earlier | Who was surveyed 1,300+ legal professionals, mostly solo and 2–5-lawyer firms | Fieldwork Sept. 19–Oct. 18, 2025 |
| Firms providing general-purpose AI tools (8am) | Reported use 46% 58% of firms with 20+ lawyers | Who was surveyed Same 8am sample | Fieldwork Sept. 19–Oct. 18, 2025 |
| Firms providing legal-specific AI tools (8am) | Reported use 34% 21% a year earlier | Who was surveyed Same 8am sample | Fieldwork Sept. 19–Oct. 18, 2025 |
| Solo and small practices using AI (Clio) | Reported use 71%–78% | Who was surveyed 1,702 U.S. legal professionals | Fieldwork Not stated |
| Midsize firms, 20–199 employees, using AI (Clio) | Reported use 86% | Who was surveyed Same Clio sample | Fieldwork Not stated |
| Midsize firms that formally adopted generative AI (SurePoint) | Reported use 63% | Who was surveyed Sample not stated | Fieldwork Data through January 2026 |
| Federal judges using at least one AI tool | Reported use 61.6% 5.4% daily | Who was surveyed 112 of 502 sampled judges | Fieldwork Dec. 1–19, 2025 |
AI use and reported gains, by Clio firm size
Shares of respondents in Clio's survey of 1,702 U.S. legal professionals (500 from a panel, 1,202 Clio customers), from its midsize-firm report released March 9. Clio gives ranges for solo and small practices.
| Firm size | Use AI | AI improved revenue | AI lets them take on more work |
|---|---|---|---|
| Solo and small | Use AI 71%–78% | AI improved revenue 31%–33% |
AI lets them take on more work
43%
Solo practices only |
| Midsize (20–199 employees) |
Use AI
86%
|
AI improved revenue
39%
|
AI lets them take on more work
65%
|
| Enterprise (200+ employees) | Use AI About the same as midsize |
AI improved revenue
59%
|
AI lets them take on more work — No comparable figure |
Revenue change is self-reported and may have causes other than AI.
Solo and small firms
Saved time becomes unpriced capacity
In Clio's survey, only 31% to 33% of solo and small firms say AI has improved revenue, and 43% to 57% operate without AI guidelines. Clients are not forcing change: in 8am's survey, 6% say clients ask for AI-related price reductions and 3% are often asked for proof of AI efficiency. Among 8am's AI users, 38% save one to five hours a week, 14% six to 10 and 9% 11 or more.
Midsize firms
High use, limited infrastructure
In Clio's survey, 65% of midsize-firm respondents say AI lets them take on more work, against 43% of solos. But only 57% have moved their data and operations to the cloud, and 30% struggle to fit AI into existing workflows. SurePoint found 81% of firm leaders report internal concern about reliability and risk.
Large firms
Gains landed in rates and margin
In Thomson Reuters' data, the Am Law 100 average standard rate passed $1,000 an hour in 2025, against about $600 at other firms, and law-firm profits grew 14.1% at a 40.7% average margin. Technology spending rose 9.7% and knowledge-management spending 10.5%, against 8.2% for lawyer compensation.
7.3%Worked-rate growth at the average U.S. firm, 2025
About 90%Share of legal dollars billed hourly
In-house and corporate buyers
Efficiency expected, AI not required
More than 80% of senior corporate counsel do not require outside firms to use AI, Thomson Reuters and Georgetown found, while corporate spending expectations are at their lowest since 2020 and one in four buyers has never had excellent value at a premium price. Legal departments are also direct buyers: the Anthropic plug-in is built for them.
Courts
Judges rely most on Westlaw's AI
In a random-sample survey of federal judges fielded in December, Westlaw's AI-Assisted Research or Deep Research was the most-used AI tool (38.4%), ahead of ChatGPT (28.6%), and 1.8% use AI to draft orders or opinions. Of chambers, 20.4% prohibit AI and 24.1% have no policy; 45.5% of judges have had no training.
Alternative providers
North America lags on alternative arrangements
Of North American lawyers, 27% report an arrangement with an alternative legal service provider or a nontraditional division, against 76% in the U.K., Europe and Australia, according to Thomson Reuters and Georgetown.
Economics
Where efficiency gains land depends on how a firm charges. At large hourly firms they have landed in rates and margin; at solo and small firms, as capacity that is not yet priced. Nearly a third of the firm leaders in SurePoint's report expect AI to reshape billing models, but none had changed theirs. Clients have not forced the question, so the pressure in the quarter came from rate scrutiny and flat spending expectations. Thomson Reuters and Georgetown projected a steep drop-off in growth going into 2026, with contraction by midyear.
Reported results: time per case, contract throughput and practitioner capacity
The quarter's results came from three kinds of source: company financial reporting, vendor and customer statements at launch, and practitioner surveys. Each holds under narrow conditions, and none is an independent measurement of a deployed legal AI system.
Company-reported
AI products grew inside incumbents' subscriptions
Thomson Reuters' generative-AI share of annual contract value rose from 24% the previous quarter and 15% in the third quarter of 2024. RELX said Lexis+ AI's enterprise-wide subscription customers more than doubled in a year, with users in the multiple hundreds of thousands. Both measures are company-defined.
28%Generative-AI-enabled share of Thomson Reuters' annual contract value, fourth quarter of 2025
Personal injury
Time saved per case in claims administration
EvenUp and its customer Jeffcoat gave the savings below at launch: EvenUp for its agents' work on calls, records requests and balance checks in early deployments, and Jeffcoat, a personal injury firm, for time once spent on the phone verifying medical balances.
9+ hoursReclaimed per case, EvenUp's estimate
3 hoursSaved per case on balance verification, Jeffcoat
AI-native firm
Contract review time at General Legal
In a March 31 interview, General Legal, which reviews commercial contracts for fixed fees with AI doing the first pass, gave the throughput below and claimed 40% margins. The figures are the firm's own, for standard commercial agreements.
About 2.2 hoursMaster services agreement review, down from 8–10
300–600Contracts per lawyer a year
Practitioner surveys
Firms report more work, not lower prices
Of Clio's midsize-firm respondents, 65% say AI lets them take on more work, 58% say it enabled more complex work and 44% report better client satisfaction. In 8am's survey, 6% of AI users report no productivity benefit, down from 16% a year earlier, and 33% report better quality without hours saved.
Benchmark
Agents completed fewer than a quarter of professional tasks
On Mercor's APEX-Agents benchmark, published Jan. 21, AI agents completed fewer than 25% of simulated banking, consulting and law tasks. Mercor, a vendor, built the benchmark.
Cross-industry context
Broad return surveys measure finance, not legal work
PwC's January CEO survey found 56% of 4,454 chief executives had seen no significant financial benefit from AI, and a March revision of an NBER working paper reported that 89% of firms, in a survey of about 6,000 executives, saw no productivity impact. Both are cross-industry self-reports; neither measures the accuracy or speed of legal work.
What the results show
Each favorable result holds for a defined task under the vendor's or firm's own measurement: claims administration in personal injury, standard commercial contracts, firms' estimates of their capacity. None establishes accuracy across legal work. A firm that wants a comparable figure has to measure time and error rates on its own matters, before and after.
Four claims from the quarter that need correcting
Several widely repeated statements from the quarter carry a distinction that changes what they mean for a buyer or a litigator.
Funding · Overstated
A $720 million loan made up 30.79% of the $2.34 billion total
Claim under review: Legal tech raised $2.3 billion in the first quarter.
Legal Complex's $2.34 billion across 103 deals includes Relativity's seven-year, JPMorgan-led term loan: debt, part of which refinances an existing private credit facility, not equity raised from investors. Without it, the Legal Complex total would be about $1.62 billion. Legal Technology Hub, which counts a narrower set of deals, recorded $1.42 billion in 35 rounds.
Products · Mischaracterized
The plug-in is an in-house contract workflow with no legal-research connector
Claim under review: Anthropic launched a rival to Westlaw and Lexis.
It reviews contracts and triages NDAs against a team's own playbook and connects to office and storage systems, not to case law. Ken Crutchfield, a former executive at LexisNexis, Thomson Reuters and Wolters Kluwer, noted that its skills are transactional while incumbents earn mainly from litigation, research and compliance. On Feb. 25, LexisNexis said it would integrate the plug-in into Protégé.
Usage · Misread
The figure counts CoCounsel customers in 107 countries, for a product sold well beyond law
Claim under review: CoCounsel has 1 million lawyer users.
Thomson Reuters' milestone, reported Feb. 24, counts 1 million CoCounsel customers across 107 countries and territories without defining the unit, for a product it now sells across legal, risk, compliance, tax, accounting, audit and trade work. It does not show how many lawyers use the product.
Sanctions · Needs precision
The court sanctioned fabricated citations without deciding whether AI produced them
Claim under review: The 6th Circuit's March sanctions punished lawyers for using AI.
In Whiting v. City of Athens, the lawyers refused to answer the court's questions, including about AI use. The panel treated the fake and misrepresented citations as sanctionable however they were produced, so the duty attaches to what is filed, whatever tool drafted it.
Failures in the quarter: unverified citations, a vendor breach, exploitable agents and consumer harm
The quarter's legal failures were of three kinds: lawyers filing authorities they had not checked; security weaknesses at a legal-information vendor and in the agent tools legal teams were adopting; and consumer harm under alternative ownership. None was a documented failed deployment of a legal AI product.
Sanctions · March 13
Steep sanctions for fake citations on appeal
In Whiting v. City of Athens, the 6th Circuit fined attorneys Van R. Irion and Russ Egli, made them jointly liable for the appellees' fees across three appeals, imposed double costs and referred both for possible discipline. Their briefs contained more than two dozen fake or misrepresented citations, including nonexistent cases and fabricated quotations. Egli was censured in 2017 for lack of candor, and Irion was suspended from a federal district court in 2025 for lying to it.
$15,000Fine per lawyer
Court filings
AI errors in US filings rose more than tenfold in 2025
SurePoint, a legal-software vendor, citing Damien Charlotin's database of AI hallucination cases, reported 487 AI error or hallucination instances in U.S. court documents in 2025, more than 10 times the 2024 total, with licensed attorneys responsible for 37.8%.
Vendor security · Confirmed March 4
LexisNexis breach exposed legacy customer data
LexisNexis confirmed that an unauthorized party reached a limited number of servers holding mostly pre-2020 data: customer names, user IDs, business contacts, products used, surveys and support tickets. It says client and matter data, search queries, contracts and active passwords were not involved. A group calling itself FulcrumSec claims more than 3.9 million records, including data on federal judges; those claims are unverified.
Agent security
Agents with file access can be turned against their users
On Jan. 15, a proof of concept showed a prompt injection, instructions hidden in content the agent reads, making Cowork send files to an outsider. On Jan. 20, The Register reported flaws in Anthropic's Git MCP server that allowed remote code execution; they were fixed in December 2025. The legal plug-in relies on the same kind of agent and connectors, with access to a team's files.
Consumer protection
Arizona investigation found harm under nonlawyer ownership
An Arizona Republic series in February reported that more than a dozen licensed alternative business structures were accused of harming clients or breaking consumer-protection laws. At least half do business out of state, some operate as call centers farming cases nationally, a nationally advertised personal injury licensee is owned by a nonlawyer with no Arizona connection, and only two firms had been mildly disciplined.
Who is accountable
In each case responsibility stayed with an identifiable person or organization: the filer for citations, the vendor for its servers, the firm for which files an agent can open, and the licensee and its regulator for consumer outcomes. Only 9% of firms in 8am's survey enforce a written AI policy and 54% have no training plans, so most cannot yet show who checked an AI-assisted document.
Courts set verification duties; ownership and referral models drew scrutiny
The quarter's legal-specific rule changes came from courts and court regulators rather than legislatures. A federal district court wrote an enforceable verification duty, Arizona's court regulators proposed tightening the main U.S. route to nonlawyer ownership, and the Solicitors Regulation Authority for England and Wales authorized a second technology-led firm. Cross-sector AI laws continued on their published schedules.
Court order · In effect from Jan. 28
District of Kansas: verify everything filed with AI help
Standing Order 26-01, signed by Chief Judge John W. Broomes, covers all lawyers and parties, including those without counsel. Filers must verify citations, quotations, paraphrases, analysis, facts and procedure in AI-assisted filings. The court may strike unverified filings on its own motion, then sanction, refer for discipline, disqualify, restrict filings or dismiss, and may require a sworn statement naming the tool, the AI-generated portions and personal checks of each authority.
Ownership · Proposed Feb. 10
Arizona moved to tighten its ABS program
After the Arizona Republic findings, court regulators on Feb. 10 backed tighter rules for alternative business structures, including limits on call-center firms and a draft requirement that licensees benefit Arizonans at least in part. The proposals were not adopted by March 31. A Stanford study counted 136 licensed entities as of April 2025.
UK authorization · February
A second technology-led firm authorized in England and Wales
The SRA authorized LawFairy, which assesses U.K. immigration eligibility with a deterministic rules engine whose logic lawyers validate, uses no large language model and charges individuals £149 per report. It is the second technology-led firm authorized after Garfield.Law in 2025.
Referrals
AI claims platforms paid through attorney lead fees
Mighty is free to consumers and earns only when a user hires a network attorney, who pays a lead-generation fee; Mighty says its software exercises no legal judgment and so is not practicing law, a question regulators have not resolved. AlphaLit raised a $3.2 million seed round on Jan. 13 to score small cases and route them to lawyers. Both rest on the mechanism behind Arizona's call-center findings: paying for, or profiting from, routing potential clients.
Insurance
General policies excluded AI; one lawyers' insurer distributed a tool
ISO generative-AI exclusions for commercial general liability policies became available in January, and carriers that had sought approval for AI exclusions in late 2025 were moving ahead with them. On Feb. 25, Wisconsin Lawyers Mutual Insurance Company, a lawyers' professional liability insurer, said it would offer policyholders a discounted subscription to Querious, an AI tool that prompts on legal issues, summarizes meetings and drafts billing entries.
Confidentiality
Consumer chatbot logs can be ordered produced
Around Jan. 5, Judge Stein affirmed an order requiring OpenAI to produce 20 million de-identified ChatGPT logs in copyright litigation. For lawyers who use consumer tools on client matters, the ruling shows that chat histories can be compelled in litigation.
AI rules and dates in force or scheduled at the end of the quarter
Dates as set in law, orders or published proposals by March 31, 2026.
| Rule or order | Date | Status at March 31 |
|---|---|---|
| California SB 53 and related laws; Texas TRAIGA; Illinois HB 3773 | Date Jan. 1, 2026 | Status at March 31 In effect |
| District of Kansas Standing Order 26-01 | Date Jan. 28, 2026 | Status at March 31 In effect until further order |
| Arizona ABS rule tightening | Date Proposed Feb. 10, 2026 | Status at March 31 Not adopted in the quarter |
| White House National Policy Framework for AI | Date Released March 20, 2026 | Status at March 31 Nonbinding; proposes preempting some state AI laws |
| Colorado AI law (SB 24-205) | Date June 30, 2026 | Status at March 31 Scheduled; replacement draft agreed March 17, not enacted |
| EU AI Act high-risk obligations | Date Aug. 2, 2026 | Status at March 31 Scheduled in law; delay proposed Council and Parliament positions would move Annex III duties to Dec. 2, 2027 |
| California SB 942 | Date Aug. 2, 2026 | Status at March 31 Scheduled |
| New York RAISE Act, as amended | Date Jan. 1, 2027 | Status at March 31 Amendment signed March 27, 2026 |
What it changes
For U.S. litigators the binding change was local and court-made, so a firm's verification standard has to meet the strictest court it appears in. For consumer legal businesses, the open question is who may pay for, or profit from, routing potential clients. Most AI statutes that touch legal work were still scheduled or proposed, and the EU's August high-risk date could move.
Where courts, incumbents, buyers, vendors and researchers agree and differ
Positions converge on two points. Verification stays with whoever files or signs: the District of Kansas order, the 6th Circuit, Anthropic's own disclaimer and Thomson Reuters' chief executive all put it there. Value sits in the context around the model, whether curated content, playbooks or ethical walls. Views diverge on who captures that value and on whether February's selloff was rational. Crutchfield, Hasker and Engstrom say core incumbent revenue is not at risk, and JPMorgan's Mark Murphy called the wider software selloff an illogical leap; Ambrogi sees an early structural shift; Kennedy puts the threat on growth rather than survival. They are describing different revenue lines: research and litigation on one side, in-house transactional work on the other. On Arizona, the chief justice, the attorney general and a Stanford researcher answer different questions about the same program. Most named voices are vendors or investors; corporate buyers appear mainly through survey data.
Courts and regulators
John W. Broomes
Chief Judge, U.S. District Court for the District of Kansas
Lawyers and parties who use AI remain responsible for reviewing and verifying everything they file, and the court may demand a sworn account of how AI was used.
His order applies equally to self-represented parties and lets the court act on its own motion.
Court order
6th Circuit panel (Judges Stranch, Bush and Murphy; opinion by Judge John K. Bush)
U.S. Court of Appeals for the 6th Circuit
Fabricated and misrepresented citations are sanctionable however they were produced, and the court did not need to decide whether AI was used.
The penalty fell on the lawyers, not on any tool.
Appellate opinion, as reported by LawSites
Ann Timmer
Chief Justice, Arizona Supreme Court
Defended Arizona's alternative business structure program against calls for heavier monitoring.
Her court's regulators proposed tighter rules on Feb. 10.
Reported by the Arizona Republic and LawSites
Kris Mayes
Arizona Attorney General
Said the allegations raised serious questions and called for a conversation about oversight of the program.
Reported by the Arizona Republic and LawSites
Legal-information incumbents
Steve Hasker
Chief executive, Thomson Reuters
A general-purpose model cannot do what professional-grade AI can do, and legal research has seen no new entrants of any importance; in late February he said professionals are no longer deciding whether to use AI but which AI to trust.
Within three weeks of the call, Thomson Reuters said its next CoCounsel Legal is built on Anthropic's Claude.
Earnings call, third-party transcript; company event via LawSites
Erik Engstrom
Chief executive, RELX
RELX sees workflow software vendors as complements rather than competitors and, asked about tools such as Cowork, said it will not license its core content, the centerpiece of its strategy.
On Feb. 25, LexisNexis said it would integrate the plug-in into Protégé, grounded in Lexis content.
Results presentation, Feb. 12
Law-firm buyers
Michael Kennedy
Head of innovation and legal technology R&D, Addleshaw Goddard
Incumbents will not disappear because of a plug-in, which is probably more a bet against their future growth than their survival, and a generic per-user contract plug-in is now a harder sell.
Addleshaw Goddard buys legal AI tools; his per-user example was a £10-a-seat contract plug-in.
Law-firm buyer, via the Law Society Gazette
Dwuan Hammond
Chief financial officer, Jeffcoat Injury & Car Accident Lawyers
In personal injury, the time a case sits on the desk is everything; agents that verify medical balances move cases to disbursement faster.
His firm reports saving three hours per case on balance verification.
EvenUp customer, via LawSites
AI-native vendors and investors
Winston Weinberg
Chief executive and co-founder, Harvey
“It's becoming the system through which legal work gets done”
Incumbents dispute this position: that the work itself, not only research, moves onto an AI platform.
Vendor, funding announcement
Max Junestrand
Chief executive and co-founder, Legora
“the pace of adoption in the U.S. has exceeded our expectations”
He cites a 9-to-1 ratio of U.S. to European legal spending to explain Legora's U.S. push.
Vendor, funding announcement
Julien Bek
Sequoia
About $60 billion of externally handled U.S. legal work, including about $36 billion of paralegal and legal process outsourcing work and $20 billion to $25 billion of transactional work, could be absorbed by AI autopilots.
Sequoia co-led Harvey's March round.
Investor essay, via Artificial Lawyer
Independent analysts and researchers
Ken Crutchfield
Spring Forward Consulting; former LexisNexis, Thomson Reuters and Wolters Kluwer executive
“Cowork's legal skills skew toward transactional and drafting”
He argues the market overreacted because incumbents earn mainly from litigation, research and compliance.
Consultant, guest post on LawSites
Bob Ambrogi
Legal technology journalist, LawSites
“But for legal tech, it is landing more like a tsunami than a drop.”
He also notes that vendors keep advantages in data and expertise.
Journalist, LawSites
Mark Murphy
Head of U.S. enterprise software research, JPMorgan
Called it an illogical leap to suggest that a new plug-in could replace every layer of mission-critical enterprise software.
Sell-side analyst, via Reuters
Natalie Knowlton
Stanford
Has seen no data showing that ABS firms harm consumers at a higher rate.
The Arizona findings document complaints, not comparative harm rates.
Researcher, via LawSites
Three decisions for legal leaders
Managing partner or chief operating officer, law firm
Decide where AI efficiency goes: rates, capacity or price
At large firms the gains have so far landed in rates and margin, with about 90% of dollars billed hourly; at small firms they appear as capacity. Clients are not yet demanding AI discounts, but corporate spending expectations are at their lowest since 2020, and an AI-native firm now quotes about $500 for a contract review.
Which matter types will we price per matter before a client or competitor sets the price, and how will we show clients what AI changed in the work?
General counsel
Decide which first-pass contract work moves in-house
The Anthropic plug-in and Harvey's in-house deployments let a legal department triage NDAs and review contracts against its own playbook. More than 80% of senior counsel still do not require outside firms to use AI.
Which contract types will our team handle on its own playbook, and what will we then expect outside counsel to charge for the rest?
Chief information officer or head of legal technology
Decide which systems an agent may read and act in
Agents now reach email, document stores and e-signature through standard connectors. The quarter brought a demonstrated prompt-injection attack that made a desktop agent send out files and a breach at a legal-information vendor. Harvey and Intapp extended ethical walls to AI access, and only 57% of midsize firms have moved their data and operations to the cloud.
For each AI tool, which repositories, mailboxes and matter systems can it reach, do our ethical walls apply to it, and is every action logged?
Four actions to take now
1
Make the Kansas verification record a firm standard
For every AI-assisted filing, record the tool used, the AI-generated portions and who checked each authority, the details the District of Kansas order lets a court demand. The firm can then answer a court's questions without reconstructing the work.
Litigation and risk partners
2
Write the playbook down as a portable file
Capture positions on liability caps, indemnities, governing law and NDA defaults in a plain document any tool can use, as the Anthropic plug-in does. The firm or department then owns its most valuable context and can change vendors without rebuilding it.
Knowledge management or legal operations
3
Inventory agent access and extend walls and logs
List each AI tool's connectors and file permissions, apply ethical-wall policies to them and turn on action logging. The result is a map of which tools can reach which client data, before a prompt-injection attack or a vendor incident tests it.
Chief information or security officer
4
Measure the cost of standard matters, review included
Time two or three rules-based matter types, such as NDAs and standard commercial contracts, from intake to the lawyer's sign-off, with and without AI. The firm gets a per-matter cost to set against its hourly bill and against AI-native prices of about $500 per contract review and $250 per NDA.
Finance and pricing
References
Covers Jan. 1 through March 31, 2026, using sources published within 30 days of the quarter's end. Several surveys cited were fielded in late 2025 and published during the quarter; the funding tallies were published in April.
Capability and cost
- knowledge-work-plugins: legal (Legal Productivity Plugin) — README, CONNECTORS and LICENSE at commit 307533e (2026-03-18) — Anthropic (GitHub), 2026-01-29
- Claude Opus 4.6 — Anthropic, 2026-02-05
- Introducing Claude Sonnet 4.6 — Anthropic, 2026-02-17
- Gemini 3.1 Pro on Gemini CLI, Gemini Enterprise and Vertex AI — Google Cloud, 2026-02-19
- Introducing GPT-5.4 — OpenAI, 2026-03-05
- Gemini 3.1 Flash-Lite: Built for intelligence at scale — Google, 2026-03-03
Voices and commentary
- Anthropic's Legal Plugin for Claude Cowork May Be the Opening Salvo In A Competition Between Foundation Models and Legal Tech Incumbents — LawSites, 2026-02-03
- Thomson Reuters (TRI) Q4 2025 Earnings Transcript — The Motley Fool, 2026-02-05
- Guest Post: Ken Crutchfield on 10 Reasons the Market Overreacted to Claude's Legal Skills Announcement — LawSites (guest post by Ken Crutchfield), 2026-02-05
- SaaSpocalypse now — Law Society Gazette (Joanna Goodman)
- Autopilots Can Absorb $60bn of Legal Work – Sequoia — Artificial Lawyer, 2026-03-30
Products and delivery
- LexisNexis Embraces Anthropic Claude Cowork Legal Plugin — Artificial Lawyer, 2026-02-25
- Thomson Reuters, Anthropic + A Surprise Video — Artificial Lawyer, 2026-02-25
- Three Years After Launching As First AI Legal Assistant, CoCounsel Reaches 1 Million Users and Thomson Reuters Teases What's Ahead — LawSites, 2026-02-24
- LexisNexis Launches Lexis+ with Protégé, Replacing Lexis+ AI with an End-to-End Workflow Platform — LawSites, 2026-02-24
- Exclusive: Smokeball and Thomson Reuters Partner to Integrate CoCounsel Legal AI with Practice Management Platform — LawSites, 2026-03-25
- Centerbase Launches Native ndMAX Integration, Connecting Practice Management to AI-Powered Document Workflows — LawSites, 2026-03-24
- EvenUp Launches AI Communication Agents to Handle Routine Tasks in PI Cases; Also Enhances Its AI Drafting — LawSites, 2026-01-28
- Can AI Agents Settle PI Cases? Mighty Says They Can and Have, As It Launches Free, Direct-to-Consumer Platform to Resolve Accident Claims — LawSites, 2026-01-28
- How Do AI-Native Law Firms Work? — Artificial Lawyer, 2026-03-31
- Relativity Launches aiR for Case Strategy to Extract and Organize Key Facts Using Generative AI — LawSites, 2026-01-12
- LexisNexis Launches Preview of Protégé AI Workflows with Hundreds of Pre-Built Tools and Custom Workflow Builder — LawSites, 2026-01-21
- DISCO Launches Scaled Agentic AI Tool for Large Discovery and Fact Investigation Matters — LawSites, 2026-02-09
- Harvey Partners With Intapp For 'Ethical Wall Enforcement' — Artificial Lawyer, 2026-02-23
- American Arbitration Association Launches Resolution Simulator, Expanding Its AI Arbitrator Tool — LawSites, 2026-03-04
Capital and markets
- Legora raises $550 million Series D to fuel US growth — Legora, 2026-03-10
- Harvey Raises at $11 Billion Valuation to Scale Agents Across Law Firms and Enterprises — Harvey, 2026-03-25
- Legora, Harvey Lead Subpar Q1 Of Legal Tech Funding — Law360, 2026-04-03
- Legal Tech Raised $2.3B in Q1 '26, But 3 Companies Dominate — Artificial Lawyer (data: Legal Complex), 2026-04-13
- Legal Tech Funding: 2026 Is On Track to Outpace 2025 — Legal Technology Hub, 2026-04-14
- Record-breaking funding: AI global Q1 2026 — Crunchbase News, 2026-04-01
- Anthropic raises $30 billion Series G funding at $380 billion post-money valuation — Anthropic, 2026-02-12
- Orbital Raises $60M Series B to Automate Real Estate Law with AI and Fuel U.S. Expansion — LawSites, 2026-01-26
- Steno Secures $49M Series C to Fuel Rapid Expansion and the Next Evolution of Transcript Genius — LawSites, 2026-03-30
- Harvey confirms $11B valuation: Sequoia triples down — TechCrunch, 2026-03-25
- Legora reaches $5.55 billion valuation as AI legaltech boom endures — TechCrunch, 2026-03-10
- Anthropic's new AI tools deepen selloff in data analytics and software stocks, investors say — Reuters via Yahoo Finance, 2026-02-04
- Anthropic legal tool jolts share price of global legal data leaders — Canadian Lawyer, 2026-02-03
- Selloff wipes out nearly $1 trillion from software and services stocks as investors debate AI's existential threat — Reuters via Profit (Pakistan Today), 2026-02-05
- Data Intelligence Company Relativity Confidentially Files for IPO; Would Be First In Legal Tech Since 2021 — LawSites, 2026-03-23
- Alexi Fires Back at Fastcase Lawsuit with Counterclaims Alleging Anticompetitive Conduct Following Clio's $1B Acquisition — LawSites, 2026-01-18
- In A Marriage Of Legal Editing Tools, BriefCatch Has Acquired WordRake and Its 12 Editing Patents — LawSites, 2026-02-11
- LawNext: From Customer to Acquirer: Filevine's Ryan Anderson and Pincites' Sona Sulakian on Building AI Contract Intelligence — LawSites, 2026-01-29
- Sirion Completes Majority Investment from Haveli, Aiming to Accelerate AI Push in CLM Market — LawSites, 2026-02-23
- JPMorgan-led refinancing cuts Relativity's private debt costs — Private Equity Wire (restating Bloomberg), 2026-01-16
- AI Startup AlphaLit Raises $3.2M Seed Round To Screen and Score Smaller Cases and Route them to Lawyers — LawSites, 2026-01-13
Market and practice
- Thomson Reuters Reports Fourth-Quarter and Full-Year 2025 Results — Thomson Reuters, 2026-02-05
- RELX 2025 results presentation transcript — RELX, 2026-02-12
- AI Adoption Among Legal Professionals Has More Than Doubled in a Year, New 8am Report Finds, But Firms Lag Far Behind Individual Practitioners — LawSites, 2026-03-05
- 2026 Legal Trends for Mid-Sized Law Firms — Clio, 2026-03-09
- Legal Industry Reaches AI Tipping Point: Majority of Lawyers Now Using Gen AI Despite Persistent Reliability Concerns — LawSites, 2026-03-25
- Survey Finds Majority of Federal Judges Have Used AI in Their Work, But Daily Use Remains Rare — LawSites, 2026-03-30
- 2026 Report on the State of the US Legal Market — Thomson Reuters Institute and Georgetown Law Center on Ethics and the Legal Profession, 2026-01-07
- 2025 Q4 Law Firm Financial Index — Thomson Reuters Institute, 2026-02-10
Courts and regulation
- LawFairy 'Technology-Only Law Firm' Gets Regulatory Approval — Artificial Lawyer, 2026-02-24
- Sixth Circuit Slaps Steep Sanctions on Two Lawyers for Fake Citations and Misrepresentations in Appellate Briefs — LawSites, 2026-03-18
- Standing Order 26-01, In re: Use of Artificial Intelligence ("AI") in Preparing Court Filings — US District Court for the District of Kansas, 2026-01-28
- Arizona Republic Investigation Finds Consumer Harm, Loopholes, and Conflicts of Interest in Arizona's Legal Regulatory Reform — LawSites, 2026-02-20
- Verisk to Roll Out New General Liability Exclusions for Generative AI Exposures — Independent Insurance Agents & Brokers of America (Independent Agent), 2025-10-21
- Insurance carriers quietly back away from covering AI outputs — CSO Online, 2026-04-16
- Querious Collaborates with Wisconsin Lawyers Mutual Insurance Company to Bring Risk Management and Productivity Tool to Policy Holders — LawSites (press release), 2026-02-26
- ChatGPT creator must turn over 20M chat logs in copyright litigation, federal judge says — ABA Journal, 2026-01-08
- New state AI laws are effective on January 1, 2026, but a new executive order signals disruption — King & Spalding, 2025-12-29
- White House releases the National Policy Framework for AI: key points — DLA Piper, 2026-03-23
- Artificial intelligence working group agrees on framework to replace Colorado law — Colorado Politics, 2026-03-17
- EU Digital Omnibus on AI update: Council and Parliament agreed positions — Addleshaw Goddard, 2026-03-27
- New York's Frontier AI Law Gets a California Makeover - With Some Key Differences — Cooley, 2026-04-01
Research and outcomes
- Introducing APEX-Agents — Mercor, 2026-01-21
- PwC 2026 Global CEO Survey (29th) — PwC, 2026-01-19
- Firm Data on AI (NBER Working Paper 34836), March 2026 revision — NBER / authors (Yotzov, Barrero, Bloom, Davis et al.), 2026-03-19
- LexisNexis Says Data Breach Has Been Contained; Hackers Claim Access to Government and Law Firm User Data — LawSites, 2026-03-04
- Contagious Claude Code bug Anthropic ignored promptly spreads to Cowork — The Register, 2026-01-15
- Anthropic quietly fixed flaws in its Git MCP server that allowed for remote code execution — The Register, 2026-01-20
Thoughtful engineering for what’s next.
Have a question about what this means for your team? Let’s talk it through.
Start a conversation