# For the AI giants, law is not a market: it is an enterprise vertical

> Reference HTML page: https://leaid.ai/en/articles/geants-ia-marche-droit-verticale
> Language: en
> Author: Mamadou Waggeh, Founder of Leaid
> Date: 2026-07-15

In under six months, Anthropic, Microsoft, Perplexity, OpenAI and Google have all staked out positions in the legal market. Our reading: for these players, legal is not a destination but one of the verticals of a broader ambition — to become the software infrastructure of the enterprise. Legaltechs are entering an ambivalent relationship with suppliers on which they depend and which are courting their own clients; French and European champions are already in the crosshairs. For legal professionals, the conclusion is operational: multi-solution architecture and interoperability must become market standards.

## Introduction

In under six months, Anthropic, Microsoft, Perplexity, OpenAI and Google have all staked out positions in the legal market. Our reading is simple: for these players, legal is not a destination. It is one of the verticals of a much broader ambition — to become the software infrastructure of the enterprise, that is, the environment in which all knowledge work is organised. This lens changes everything: how their announcements should be interpreted, the position of French and European legaltechs, and the architecture decisions that legal professionals must take now.

The running tally kept by Law.com shows the scale of the movement. In late January 2026, Anthropic launched a legal plugin in Claude Cowork; in April, Microsoft unveiled its Legal Agent while Freshfields signed a multi-year agreement giving nearly 6,000 lawyers and staff access to Claude; in late June, Perplexity announced Computer for Counsel; at the same time, OpenAI hired the co-founder of Ironclad to build its legal vertical, while Google’s funds stepped up their investments in legaltechs. As Forbes noted in late June, the question is no longer whether model providers will move into law, but how intensely they will fight over the market.

Yet among the leaders of law firms, in-house legal departments and legaltechs, we still observe a largely defensive reading of these announcements: one more competitor, one more tool, a valuation hit to absorb. This article offers a different reading, focused on the real strategy of these players, on the structural dilemma in which it places legaltechs, and on its very concrete consequences for the French and European ecosystem.

## 1. Six months that reshaped the market: what the AI giants actually launched

To reason soundly, the facts must first be established. The overview below, drawn from Law.com’s tracking, distinguishes what each player has actually brought to market — a product, an ecosystem, capital or a team.

- **Anthropic**: a legal plugin in Claude Cowork in late January 2026 (contract review, risk detection, compliance workflows), followed three months later by twelve plugins by practice area — commercial, employment, intellectual property, litigation, AI governance, regulatory — and by integrations with more than twenty legaltechs via MCP connectors: Docusign, Ironclad, Relativity, Everlaw, iManage, NetDocuments, Thomson Reuters and Harvey. Add to that a Microsoft 365 integration and law-firm deployments: Freshfields (multi-year agreement, nearly 6,000 users across 33 offices, early access to future models) and Hanson Bridgett (200 lawyers).

- **Microsoft**: ubiquitous in law firms since 2020 through Microsoft 365, Azure and Teams, the software vendor launched its first specifically legal tool in April, a Legal Agent embedded in Word — document analysis, suggested edits, compliance checks against the organisation’s internal standards — built by hiring teams from the former legaltech Robin AI. Its main role nonetheless remains that of an ecosystem: Litera, NetDocuments, LawToolBox and Luminance exist inside its environment, some solely as Microsoft 365 applications.

- **Perplexity**: Computer for Counsel, launched in late June for Enterprise and Max customers — NDA review, regulatory dashboards, legal research — with Docusign, NetDocuments, Box, Carta, DeepJudge and Midpage integrations, and others announced with Clio and Ironclad. A telling detail: the product grew out of internal use by Perplexity’s own legal department.

- **OpenAI**: no legal product yet, but a legal vertical under construction, entrusted in summer 2026 to Jason Boehmig, co-founder and former head of Ironclad and himself a former corporate lawyer. OpenAI is also an early investor in Harvey — whose platform runs on its models — and co-developed with it a model specialised in case law.

- **Google**: no legal offering of its own, but a strategy of capital and partnerships. Its funds GV and Gradient Ventures have invested in Harvey (Series C then Series D rounds of $100 million and $300 million), Hebbia, Laurel, Rocket Lawyer and — a decisive point for our argument — in European players: Lawhive in the United Kingdom (£60 million in February 2026), Contractbook in Sweden, Flank in Berlin. Google Cloud has also been equipping Freshfields since April 2025 (Gemini, Vertex AI, Agentspace), and the firm has built its due diligence platform on it.

Five different strategies — product, ecosystem, capital, talent, partnerships — but one and the same direction: each of these players now regards law as territory to be occupied.

## 2. Law is not their destination: the logic of the enterprise vertical

None of these companies was created to serve legal professionals. As Forbes pointed out, they were created to build ever more capable intelligence systems; as they progressed, those systems encountered medicine, finance, engineering, research — and law. For a model provider, law is not a market apart: it is one knowledge domain among others and, above all, one of the buying centres of an enterprise account. The in-house legal department carries weight in a large group’s procurement decisions; law firms are deep-pocketed, document-intensive organisations whose adoption lends the offering credibility with all the other advisory professions.

This logic must be taken seriously, because it sheds light on the real nature of the offerings. What Anthropic is selling Freshfields is not legal software: it is thousands of seats on a general-purpose platform, to which the legal plugins are the entry point. What Microsoft is adding to Word with its Legal Agent is not a legaltech: it is one more reason never to leave its environment. What Perplexity is industrialising with Computer for Counsel is the extension to its customers of a practice that originated in its own legal department. In all three cases, legal is not the end in itself: it is a vertical in the service of a platform strategy whose objective is to become the software layer on which the entire enterprise works.

This scenario had been anticipated. As early as 2020, before the generative AI wave, Mark Cohen was asking in Forbes “Law Is Not Ready For Amazon. Is Amazon Ready For Law?”: the next transformative force in law would come from organisations outside the sector, endowed with capital, data and a platform scale that bear no comparison with those of legal players. The prediction has come true — not through e-commerce, but through model providers. And one weak signal deserves attention: Hanson Bridgett says it is assessing whether Claude “eliminates the need” for certain specialised legal AI tools. When the general-purpose platform absorbs use cases one by one, the climb up the software stack is no longer a hypothesis: it is a movement already under way.

## 3. The legaltech dilemma: customers, partners and competitors of their own suppliers

For legaltechs, this configuration creates a relationship with few parallels in the history of software: their most critical suppliers are also their most serious competitors. Almost all legal AI applications rely on models from Anthropic, OpenAI or Google, consumed via API: the model is the heart of their product, and it is rented from a third party. Harvey, often presented as the category champion, is built on OpenAI’s models — OpenAI has also been its investor since 2024 and is now building its own legal vertical, entrusted to the founder of Ironclad. The supplier, the shareholder and the competitor are the same company.

The markets' reaction to Anthropic’s first announcement — up to 20% of market capitalisation lost in a week for listed legaltechs, according to Law.com — has often been described as excessive. Yet it correctly identified the structure of the problem: model providers are now pursuing the same end customers as the legaltechs they supply. When a firm like Hanson Bridgett says it is assessing whether the general-purpose platform makes certain specialised subscriptions redundant, the feared scenario is no longer an analyst’s projection: it is a client saying so.

Hence the integration dilemma. Refusing to connect to the providers' ecosystems — plugins, MCP connectors — means becoming invisible where usage is now taking shape. Connecting means strengthening the platform, commoditising one’s own differentiation and exposing one’s client relationship to a player that covets it. Most have made their choice: more than twenty legaltechs featured in Anthropic’s April announcement, and LexisNexis itself has integrated Claude’s plugins into Lexis+ Protégé. This is not naivety; it is a rational calculation in an unfavourable balance of power: each is betting that it will be the indispensable building block of the ecosystem rather than the building block absorbed by it.

We set out what remains defensible in our analysis of [the battle for control of the workflow](/en/articles/ia-juridique-maitrise-du-workflow): proprietary domain data, functional depth on specific use cases, native compliance and the relationship of trust with professions that general-purpose platforms understand poorly. It is precisely on this ground that the fate of French and European players will be decided.

## 4. France and Europe are not spectators: champions already in the crosshairs

Europe is not watching this movement from afar: it is already one of its playing fields. The European investments of Google’s funds show as much — Lawhive in the United Kingdom in February 2026, Contractbook in Sweden, Flank in Berlin. The hyperscalers' capital and partnerships have started shopping on the continent, and there is no reason the movement should stop at English-speaking or Nordic players.

French and European champions, after all, hold precisely what general-purpose platforms lack: structured continental law corpora, workflows designed for civil-law practice, intimate knowledge of the regulated professions — professional ethics, professional secrecy, the role of the professional bodies — institutional roots and native European compliance. As model providers seek to make their legal verticals credible outside the English-speaking world, these assets will become strategic. We are convinced this appeal can only grow: commercial partnerships first, then equity stakes, and finally acquisitions.

This appeal is double-edged. It validates the quality of the ecosystem — France has recognised legaltechs, solid software vendors and model providers such as Mistral AI — and it opens up market opportunities, capital and global distribution channels. But it also carries two symmetrical risks: that of seeing European champions reduced to the role of a local distribution layer for American platforms, and that of seeing them absorbed before reaching critical mass. The window in which a European player can still choose its destiny — grow, ally or sell on its own terms — is open, but it will not stay open indefinitely.

Finally, Europe’s specific characteristics change the rules of the game for everyone. The staggered application of the EU AI Act, the GDPR, professional secrecy, sovereignty requirements and public procurement rules create real friction for general-purpose platforms: they must prove data localisation, auditability, governance and contractual robustness before equipping a law firm, an institution or a public buyer. This friction is a window of opportunity for local players — on one condition, which we have already set out in these pages: treat [compliance and sovereignty](/en/articles/ia-generative-secteurs-regules-confiance) as a demonstrable product advantage, never as a protection that can be taken for granted.

## 5. For legal professionals: architecture becomes the strategic decision

For a law firm, an in-house legal department or an institution, the practical consequence of all of the above comes down to one sentence: the question is no longer “which tool to choose” but “which architecture to build”. Standardising everything on a single general-purpose platform offers immediate convenience — one contract, one interface, one deployment — at the price of structural dependence on a player for which law is only one vertical among others: its priorities, its prices and its functional scope will evolve according to considerations that owe nothing to the needs of legal professionals. Conversely, stacking specialised tools without a common core scatters data, multiplies risks and makes it impossible to measure value.

The answer we recommend is a multi-solution architecture organised around a controlled core.

- **A controlled document and governance core**: the firm’s data — matter files, document templates, internal know-how, records — lives in a system the organisation controls, never in the memory of a third-party platform.

- **Interchangeable building blocks**: general-purpose assistant, legal research, contract management, specialist tools — each function is a building block chosen for its own value and replaceable without rebuilding the whole.

- **Standard connectors**: the building blocks communicate through documented APIs and open protocols, never through exclusive proprietary integrations.

- **Overall governance**: documented authorised uses, controlled access, measured value and managed incidents — whichever tool is involved.

- **Tested reversibility**: the exit from each building block — data, prompts, workflows — is written into the contract and verified before signing, not discovered in the middle of a crisis.

This architecture is no more costly than dependence: it is more demanding at the outset and far less risky over time. Above all, it is the only one that makes it possible to extract the full value of each tool — the general-purpose assistant for the breadth of its capabilities, the specialised solutions for their domain depth — without entrusting one’s operational destiny to a single supplier.

## 6. Interoperability must become a market standard

A multi-solution architecture is only possible if the tools agree to talk to one another. That is why interoperability must stop being a selling point and become a market standard: a default requirement of buyers, verified before any signature. The success of MCP connectors offers a glimpse of this — an open protocol published by Anthropic became, within a few months, the common language of legal integrations. That is good news for interoperability, but a de facto standard controlled by one supplier is never neutral: the history of browsers and mobile operating systems has taught Europe what it costs to let a single player set the rules of the game.

And buyers — law firms, in-house legal departments, institutions — collectively hold the power to impose this standard. In concrete terms, every tender and every specification should require:

- **Open, documented connectors**: the tool exposes its functions through a public API, without punitive surcharges or arbitrary restrictions on use.

- **Full export of data and deliverables**: in usable formats, at any time, without prohibitive fees or deliberate delays.

- **Workflow portability**: prompts, templates, automations and settings must be retrievable and redeployable elsewhere.

- **No exclusivity**: no clause may prevent the tool from being combined with those of a competitor, including a rival model provider.

- **Contractual reversibility**: the exit procedure is described, costed and enforceable from the moment of signing.

[Professional bodies, in-house counsel associations](/en/expertises#organisations) and public buyers have a historic role to play here: writing these requirements into common frameworks, as the EU AI Act and the Data Act are beginning to do for cloud computing services. And for French and European legaltechs, interoperability is not a concession to the market: it is a survival strategy. In a market dominated by general-purpose platforms, the future belongs to excellent, interoperable building blocks — not to closed mini-platforms that reproduce, on a smaller scale, the model of those that threaten them.

## Conclusion

The AI giants will not be leaving the legal market, because law was never their objective: their objective is to be the infrastructure on which all knowledge work runs, and legal is a vertical too lucrative, too document-intensive and too influential as a prescriber to be left aside. Faced with this movement, fatalism is as poorly calibrated as denial. As Forbes reminded its readers, every major technology wave has ended up shifting value to the last mile: implementation, architecture, governance, training — everything that happens after the technology is bought.

That is where the game will be decided for the French and European ecosystem. The legaltechs that endure will be the best building blocks of architectures they will not fully control. The law firms and in-house legal departments that actually extract value from AI will be those that have built a reversible architecture rather than adopted a platform by default. And interoperability will become a market standard if — and only if — buyers demand it now, while the market is taking shape. In a reshaping of this magnitude, the only truly untenable position is that of the spectator.

---

Markdown version of https://leaid.ai/en/articles/geants-ia-marche-droit-verticale — Leaid · https://leaid.ai · https://leaid.ai/llms.txt