Anduril’s $61B Valuation: Why Investors Bet $5B on the Future of Warfare
On May 12, 2026, Anthropic launched a sweeping bid to become the operating system for the legal profession. Building on a February release that made legal professionals its most engaged user base, the AI developer rolled out 20 open-protocol connectors and 12 practice-specific plugins, embedding its Claude assistant directly into the legal technology stack.
The market response was nuanced. Incumbent shares registered pressure—Thomson Reuters dipped 2.1%, RELX fell 1.5%, and enterprise stalwarts DocuSign and Box saw marginal declines. Yet the reality is far more dangerous to the old guard than simple technological replacement. Anthropic is not trying to beat legal databases at their own game. It wants to make them subordinate endpoints in a Claude-controlled workflow.
The Architecture of Integration
The new release does not attempt to replicate vast repositories of case law or the defensible audit trails of enterprise discovery platforms. Instead, via Model Context Protocol (MCP) connectors, Claude reaches directly into the systems lawyers already rely on: contract lifecycle tools (DocuSign, Ironclad), document management vaults (iManage, NetDocuments), e-discovery platforms (Everlaw, Relativity), and the 31-million-document corpus of Legal Data Hunter.
Crucially, this happens where the work lives. Claude operates natively within Microsoft Word, Outlook, Excel, and PowerPoint, carrying context seamlessly across the suite. A redline negotiated in Word translates directly into an Outlook cover note without re-prompting.
Coupled with 12 plugins tailored to specific roles—from Commercial and M&A to Litigation—the system initiates with a setup interview that ingests a team’s specific playbooks and risk tolerances. But while Claude Opus 4.7 recently scored 90.9% on Harvey’s BigLaw Bench, generating a flawless summary is not the same as wielding legal authority.
The Strategic Gambit: Owning the Workflow Moment
The lazy bear thesis assumes legacy databases are doomed because Claude integrates with Westlaw. This misreads Anthropic’s strategy. Classic vertical SaaS wins by owning the data; Anthropic aims to own the agentic interaction layer.
Everlaw’s framing illustrates the dynamic perfectly: Claude serves as the natural-language interface, while Everlaw remains the governed system of record. Thomson Reuters struck a similar posture, confirming that while users access primary law within Claude, CoCounsel remains the fiduciary validation layer, neither bypassed nor replaced. LexisNexis anticipated this battlefield days earlier, launching its Protégé assistant to wrap its own authoritative content.
This is a battle over where legal work begins. Whichever platform owns the "front door" commands user attention, dictates workflow design, and ultimately seizes pricing leverage.
The Verification Tax
The most underpriced constraint in bullish AI narratives is the "verification tax." In law, time saved in drafting must be brutally adjusted for review time, risk mitigation, and professional liability.
AI hallucinations remain a litigated hazard. Judges continue to sanction attorneys for AI-generated citation errors, and on May 7, a U.S. judicial panel delayed issuing rules on AI-generated evidence, citing unresolved expert disputes.
This tax will prevent Claude from fully replacing human legal judgment, but it will not prevent massive workflow compression. Claude does not need to eliminate the lawyer to alter the economics of legal labor; it only needs to shrink the hours required for first-pass diligence.
Who Wins, Who Gets Squeezed
The immediate beneficiaries are in-house corporate legal teams. Unburdened by protecting billable hours, they seek throughput and reduced outside-counsel spend. Their adoption of Claude for routine reviews will be the clearest early indicator of economic disruption.
Incumbents like Thomson Reuters, LexisNexis, and iManage can thrive, but only if they position their proprietary governance layers as indispensable validation steps inside AI workflows.
The blast radius is expanding. Thin AI wrappers lacking proprietary data will be commoditized overnight. Low-complexity contract management modules are newly redundant. And law firms relying on billing thousands of hours for junior-associate diligence face an existential threat to their leverage models unless they restructure pricing.
The Structural Conclusion
The legal AI landscape is no longer a story of disruption by replacement. It is a contest over whose agent sits atop whose authority layer. By moving sequentially from foundation model, to desktop agent, to Microsoft 365 embedding, and now to vertical ecosystem integration, Anthropic is executing the exact playbook it deployed in financial services a week prior.
The true disruption here is the profound reallocation of control. Power is shifting away from isolated applications and junior human labor, migrating toward agentic work surfaces, verified authority layers, and attorneys skilled enough to supervise machines. For investors underwriting the future of legal technology, the defining question is no longer which model scores highest on a benchmark. The only question that matters is: who owns the workflow moment when the lawyer decides what to do next?
not investment advice
Sources: https://claude.com/blog/claude-for-the-legal-industry