When Anthropic shipped the Cowork legal plugin in February 2026, three legal-data incumbents lost roughly $285 billion in combined market cap in a single trading session. Per Canadian Lawyer's reporting, Thomson Reuters dropped -16%, RELX (LexisNexis parent) dropped -14%, and Wolters Kluwer dropped -13%. The plugin itself is open source. The product reaction wasn't about a feature — it was about a structural read on what an open, configurable legal plugin running directly on Claude does to the moats those companies built over four decades. Per LawSites' February analysis, this was the opening salvo in a foundation-model-vs-vendor competition.

What actually shipped on the Cowork launch day

Per the Anthropic legal plugin page and the knowledge-work-plugins/legal GitHub repo, the February 2026 release included two specific commands plus the configuration surface around them:

- /review-contract runs clause-by-clause review against a configured negotiation playbook with GREEN/YELLOW/RED flags plus inline redline suggestions. - /triage-nda does rapid pre-screening that categorizes incoming NDAs into standard approval, counsel review, or full review tiers. - The full plugin source is open on GitHub. Firms can fork, modify, and ship internally without an enterprise license. - Configuration is per-org playbook plus risk tolerance — not a one-size template. - Disclaimer reads "assistance not advice" — Anthropic is explicitly not positioning this as legal counsel.

The technical achievement isn't novel. Vertical legal AI vendors have shipped clause review with risk flags since 2018. What changed: the same workflow now runs directly on Claude Cowork, free, configurable, with source code published. The unbundling moved up a layer.

The $285B market reaction — what it actually priced in

Three companies took the hit on the same day: Thomson Reuters -16%, RELX -14%, Wolters Kluwer -13%. Combined enterprise value impact landed near $285 billion based on closing prices that session.

The market wasn't pricing in a single product. It was pricing in three structural reads.

First, that the contract review and NDA triage workflows — historically gated behind seat licenses ranging from quote-only enterprise contracts to per-user-month vendor deals — could now be replicated by any in-house team with Claude Cowork access. Per the pricing.csv reference data point, Anthropic's Claude Pro tier sits at $17/user/month annual or $20/user/month monthly. That's the floor for a configurable workflow that previously required vendor relationships running into five-figure annual commitments per practice area.

Second, that the open-source posture removes procurement friction. IT departments don't need to onboard a new vendor; they install a plugin. The traditional vendor sales cycle — RFP, security review, contract negotiation, deployment — gets compressed to a configuration session.

Third, the deeper read: legal data incumbents aren't competing against a model that's slightly better than theirs. They're competing against a model that ships with the workflow logic embedded plus the freedom to modify it. The Cowork legal plugin overview spoke walks through what each command actually does in practice.

The market reaction was sharp, but the structural read isn't "vendors disappear." It's that the pricing power and feature differentiation move up the stack.

Thomson Reuters responded by rebuilding CoCounsel on Anthropic — embedding Westlaw and Practical Law content into the Anthropic-powered surface. That's the converge play: vendor adds proprietary content moats on top of the foundation model, retains distribution.

LexisNexis pursued a different pattern. RELX acquired Doctrine, the French legal AI platform, on April 28, 2026 — extending the Lex/Westlaw duopoly into multilingual European markets. Different play: vendor consolidates regional content moats before foundation models commoditize the workflow layer.

The second-order pattern. Vendors with proprietary content (Westlaw, Lexis, Practical Law) retain pricing power because the model layer can run the workflow but can't generate the case law, regulatory updates, or jurisdictional commentary. Vendors whose moat was workflow-only (basic clause review, basic redlining) lose pricing power because the workflow now ships with the model.

The third-order read: the $285B reaction priced in workflow-moat erosion, not content-moat erosion. The companies whose stocks recovered fastest were the ones with proprietary content depth. The ones still flat several months later are the ones whose differentiation was UI plus workflow without proprietary research depth underneath.

In-house legal departments were the structurally happiest readers of the Cowork launch. They've spent five years trying to build internal NDA triage and clause review tooling without engineering budget. The Cowork plugin shipped that capability for the cost of a Claude subscription plus a configuration session.

The operational implication. A 50-person in-house team running on Claude Team at $25/seat/month per Anthropic's pricing gets the full Cowork legal plugin for $1,250/month total team cost. That's the floor for a workflow that vendor procurement would historically price at $50,000+ annual minimum.

The second-order effect. In-house teams that previously had to justify per-seat vendor licenses with detailed ROI memos can now stand up the same workflow inside an existing Claude deployment. Procurement battles with finance over legaltech budgets shrink — the cost is folded into a subscription the company likely already has.

The third-order effect. Vendors who sold to in-house GCs lose their easiest sale. The pitch "this is workflow you can't build in-house" stops working when the workflow ships with the foundation model. Vendors pivot toward what they CAN do that in-house teams can't replicate easily: proprietary content, regulated-industry compliance modules, multi-jurisdictional coverage. The in-house Anthropic deployment checklist breaks down what to configure first.

First-party data: how the Cowork launch surfaced in Vortex's grounding queries

In the 30 days following the February Cowork launch, Vortex's Bing AI Performance dashboard showed a measurable shift in grounding query behavior. Microsoft Copilot citations of aivortex.io that previously concentrated on "Harvey AI legal" began surfacing on "Anthropic Cowork legal" and "open source legal AI plugin" queries within two weeks of the launch.

The second-order read: when a market-moving event hits the wire, search behavior changes within 48-72 hours, but AI engine citation behavior changes faster than traditional SERP behavior. Pages already grounded for the topic earn new visibility without any new content shipped.

The third-order read: most law firms have no view of Bing AI Performance. They don't see which queries shifted, when, or which pages earn the new citations. The dashboard is free. Per the Bing Webmaster Tools setup, it takes about two weeks from registration to first useful data. Vortex was instrumented before the Cowork launch, which is why the citation shift was visible day-one. The Cowork explainer spoke is the entry point for firms that want to understand what they're missing.

Recommendations by firm size and posture

Solos and small firms (under 50 lawyers): Claude Pro at $17 annual or $20 monthly per Anthropic's pricing plus the Cowork legal plugin replaces several vendor categories at once for transactional work. Configuration takes 4-8 hours. Annual cost: roughly $240. The replacement target isn't Westlaw — it's the basic-tier contract review tools sized for small firms.

Mid-market firms (50-500 lawyers): Claude Team at $20-25/seat/month is the right tier. A 200-lawyer deployment is $4,000-5,000/month, $48,000-60,000/year — for the model layer plus the legal plugin. Add Spellbook, CoCounsel, or Harvey for proprietary content depth where the practice mix demands it. The decision isn't "replace vendors" — it's "add the foundation model layer underneath them."

BigLaw and AmLaw 100: The Cowork plugin matters less as a standalone tool and more as a signal about Anthropic's trajectory. Firms at this tier are more likely to follow the Freshfields co-build template than rely on the public plugin. The Anthropic procurement checklist for mid-market covers the mid-tier playbook; the big law next-to-go-Anthropic-deep prediction covers the BigLaw posture.

The Bottom Line: My take: The $285B market reaction priced in workflow-moat erosion, not the disappearance of legal data vendors. The vendors with proprietary content depth (Westlaw, Lexis, Practical Law) recovered fastest by converging onto the foundation model layer. The vendors whose moat was UI plus workflow lost pricing power and haven't recovered. For in-house legal ops, the Cowork plugin is the largest single procurement-budget shrink in legal AI history — workflow that previously cost five-figure annual contracts now ships at Claude subscription prices. The procurement memo writes itself.

AI-Assisted Research. This piece was researched and written with AI assistance, reviewed and edited by Manu Ayala. For deeper takes and the perspective behind the research, follow me on LinkedIn or email me directly.