Anthropic, Perplexity, and the rest of the frontier AI companies are no longer testing legal; they are shipping into it. Claude for Legal launched on May 12, 2026, with more than 20 MCP connectors and 12 practice-area plugins. Perplexity’s Computer for Counsel followed in late June 2026 as an agentic platform connecting legal data sources, document management, and workflow tools. For every existing legal tech vendor, the question has moved from defending market share to choosing which side of the new stack to stand on.
The new competitors are easy to count: Anthropic, Perplexity, and the rest of the frontier cohort. The structure they impose is harder to see. Anthropic and Perplexity were not built to serve lawyers; they were built around general intelligence, and law is one of many domains those systems now operate in.
Anthropic Shipped the Legal Stack
Anthropic launched Claude for Legal on May 12, 2026, with more than 20 MCP connectors and 12 practice-area plugins, the largest single move yet by a frontier model company into the legal market. The connectors link Claude to the software law firms and corporate legal departments already run, including Ironclad, DocuSign, Relativity, Everlaw, Consilio, iManage, NetDocuments, Definely, Box, and Datasite. The plugins are tailored to specific practice areas rather than generic workflow tooling: Commercial Legal, Corporate Legal, Employment Legal, Privacy Legal, Product Legal, Regulatory Legal, AI Governance Legal, IP Legal, Litigation Legal, plus plugins for law students, legal clinics, and a Legal Builder Hub for community-built skills.
The most loaded line on the partner list is Thomson Reuters. CoCounsel Legal, the legal AI product Thomson Reuters has rebuilt on Anthropic’s technology, is now both the substrate Claude sits on and a tool Claude can call. That bidirectional integration captures the structure of the new competition: the foundation model underwrites the application layer and increasingly competes with it. Per LawNext’s write-up of the launch, “Rather than fleeing, companies such as Harvey, Relativity, Everlaw, and Thomson Reuters are integrating deeply, betting that being part of the Claude ecosystem is better than sitting outside it.” Anthropic, valued at over $900 billion in the Artificial Lawyer coverage of the launch, is approaching the scale of the entire global legal market it is now addressing.
Anthropic also built an access-to-justice tier around the same announcement. Connectors for Courtroom5, which the company says serves roughly 80% of US civil litigants who appear without an attorney, and BoardWise, for licensed professionals navigating state board matters, are available alongside a Claude for Nonprofits discount for legal aid organizations, public defenders, and qualifying nonprofit legal services groups. The For Nonprofits tier signals that Anthropic sees legal as a domain broad enough to include both enterprise customers and the public-interest market most specialist vendors do not serve.
The scale of the release reflects Anthropic’s own data on who has been using Claude Cowork. Since the legal plugin Anthropic shipped in early February 2026 for the agentic desktop tool, legal professionals have become the most engaged Cowork users of any knowledge-work function. Per Mark Pike, Anthropic’s Associate General Counsel and product lead for legal, in Artificial Lawyer’s Q&A, legal became “the number one power-user job function in Claude Cowork, with over three times the usage of any other function.” Each plugin starts with what Anthropic describes as a setup interview that learns a team’s specific playbooks, escalation chains, risk calibration, and house style.
- More than 20 MCP connectors linking Claude to existing legal software, from Ironclad and DocuSign to Relativity and Everlaw (Anthropic, May 12, 2026)
- 12 practice-area plugins covering Commercial, Corporate, Employment, Privacy, Product, Regulatory, AI Governance, IP, and Litigation legal work
- Legal Data Hunter, a connector partner, claims a corpus of more than 31 million documents across 160-plus jurisdictions
- Courtroom5, an access-to-justice connector, serves roughly 80% of US civil litigants without an attorney
- Claude for Word, the in-document legal workflow tool, entered beta in April 2026 with legal contract review as its first example use case
What Perplexity Built This Month
Perplexity’s Computer for Counsel arrived in late June 2026 as an agentic platform built for lawyers. Per Law.com’s reporting on the launch, it integrates via MCP with DeepJudge, Docusign, NetDocuments, Carta, Box, LegalZoom, and Deel, with Clio and Ironclad integrations planned. The product is positioned as a research and administrative layer that leaves legal judgment with the lawyer, a framing consistent with Perplexity’s posture of fine-tuning models to reward factual accuracy and penalize hallucination.
Gunderson Dettmer, the Silicon Valley firm focused on the innovation economy, initially rolled out Perplexity AI firmwide in May 2025 and is now preparing to embrace Computer for Counsel’s agentic capabilities, per Law.com. Chief innovation officer Joe Green told Law.com the firm anticipates use cases in transactional research, document review for lengthier files, and context-aware workflows pulling from email, document management, and knowledge repositories via the DeepJudge connector. The Freshfields parallel is instructive: Anthropic reported roughly 500% growth in Claude usage within the first six weeks of its deployment across 33 Freshfields offices. A firmwide deployment at a top-tier venture shop is the cleanest proof point Perplexity could ask for: legal AI at scale inside a partnership whose clients include the AI vendors themselves.
The Wider Cohort Is Already Moving
Anthropic and Perplexity are the two clearest data points, but they are not alone. Per Joseph Andrew’s framing in Forbes on June 29, 2026, Microsoft continues to fold AI into enterprise legal environments, Apple is working with Google to include law in Apple Intelligence, and Google is pursuing its own legal and enterprise strategy.
The pattern that matters most is the one Anthropic’s launch made visible. Specialist legal AI vendors built on top of Claude are choosing to integrate deeply. Per LawNext, the foundation model is now “both underlining and increasingly competing with the application layer built on top of it.” That is the new shape of the market: model companies extend down into legal, application companies extend up into models, and the boundary between them is the part being redrawn. Harvey CEO Winston Weinberg told Artificial Lawyer the firm had always believed it would “end up competing with the model companies” and that the Anthropic announcement validated the original strategy.
One frontier AI vendor entering legal would be a competitive event. Three or four entering at the same time, with the same underlying logic, is a market structure shift.
A 2020 Prediction That Aged Fast
Six years before the current wave, Mark Cohen published a Forbes piece titled “Law Is Not Ready For Amazon. Is Amazon Ready For Law?” on November 23, 2020. Most readers treated it as a provocation about a single company. Looking back, the provocation was the wrong frame. The thesis underneath it held that organizations with extraordinary technology, customer relationships, capital, data, and platform scale could eventually have more influence on the future of law than many of the institutions that traditionally defined it.
Cohen was specific about which Amazon traits mattered: customer-centricity, platform, data mastery, agility, up-skilling investment, patience, and warp-speed ability to scale. He argued those traits would matter more in law than in many of the sectors Amazon had already entered, because law is fragmented, inefficient, dominated by legacy models, and sitting on a gold mine of data. His concrete prediction: an Amazon-style entrant would create a curated marketplace, accelerate access to legal services, and put pressure on non-differentiated providers.
What has arrived since March 2026 is the intelligence-vendor archetype. The companies entering legal today have model research budgets, infrastructure scale, and engineering teams that dwarf any specialist legal tech company. Per Andrew’s analysis in Forbes, the newcomers do not view law as a separate and protected market. They view law as a domain of knowledge, no different in kind from medicine, finance, or scientific research.
Cohen’s November 2020 prediction about Amazon reshaping law remains the clearest framing of what is happening now. The original disagreements about it were about whether any single non-legal company would bother to enter. The answer has arrived, and the question was never which company. It was what type of company. Intelligence vendors, with model labs and platform scale, were always the more likely entrant than a retailer. Anthropic at over $900 billion in valuation, roughly the size of the entire global legal market, is the size of entrant Cohen’s thesis was pointing toward.
Where Specialized Vendors Push Back
The doomsday framing of the shift is too neat for the vendors closest to the work. Raj Fernando and Nick Stech, co-founders of Workstorm, the Chicago-based secure collaboration platform that serves large US law firms, push back on the assumption that generic legal AI will absorb everything in its path. Their argument runs in two parts: the legal market is not generic, and the standard of care in law is not what generic AI is built for.
The standard of care and need for precision for lawyers is extremely high. What might be “good enough” or “almost perfect” for corporate professionals is not good enough for lawyers, let alone lawyers at a prestigious firm.
That position comes from a CEO with prior success in financial technology and a COO who came out of trading before moving into legal workflow automation. In a Workstorm interview on legal workflow automation, Stech described AI, particularly agentic AI, as a way to handle “deadline tracking, document routing, or risk flagging” inside firm workflows, not as a replacement for legal judgment. The Workstorm thesis is that platform vendors need to be embedded inside the firm’s existing technology stack (CRM, billing, document management) rather than sitting on top of it as a separate destination.
The argument holds together in three places. First, there is no generic law firm or generic corporate in-house team, as Fernando put it in the Forbes piece, which means frontier AI vendors who ship a single product will need to do customization work the enterprise tier does not require of most other industries. Second, law firm data stays inside the firm for security, privilege, and ethical wall reasons, which constrains how much a foundation model can learn from any one client. Third, the last-mile integration work (security review, change management, escalation chains, house style) is exactly the work frontier vendors have historically not done well. None of that defends the legacy stack, but it does argue that frontier AI vendors will need partners to land inside firms, and that the doomsday scenario assumes a level of generic substitution the Workstorm team and others do not see in the actual buying patterns of Am Law 100 firms.
The Last Mile Becomes the Battleground
Every major enterprise software wave has eventually produced a large consulting and implementation ecosystem because institutions needed help navigating change. The same pattern is showing up in legal AI. Four distinct winner categories are emerging at the same time, and the most defensible position in each is different. Implementation and consulting partners, in particular, are emerging as a distinct layer rather than a footnote.
| Winner category | What they build | Where the moat sits |
|---|---|---|
| Foundation model makers | General intelligence layer | Capital, infrastructure, engineering talent |
| Application vendors | Legal workflow products | Deep domain integration with foundation models |
| Data layer providers | Curated legal corpora | Jurisdiction-specific content at scale |
| Implementation partners | Change management and integration | Firm relationships and security clearances |
Foundation model builders compete on capital, infrastructure, and engineering talent, the dimensions Anthropic and Perplexity already dominate. Application vendors compete on how deeply they integrate with the foundation layer, which makes the Harvey-style bets on bidirectional integration a survival move as much as a partnership move. Data layer providers compete on curated, jurisdiction-specific content, and Legal Data Hunter’s more than 31 million documents across 160-plus jurisdictions is one example of how high that bar has been set. The strategic question for every existing legal tech company is which of the four layers it can credibly occupy. The most exposed position is the one that tries to stay independent of all four.
Frequently Asked Questions
What did Anthropic actually launch in legal?
Anthropic launched Claude for Legal on May 12, 2026. The release included more than 20 MCP connectors linking Claude to existing legal software (Ironclad, Relativity, Everlaw, Harvey, and others) and 12 practice-area plugins covering Commercial, Corporate, Employment, Privacy, Product, Regulatory, AI Governance, IP, and Litigation legal work, plus plugins for law students, legal clinics, and a community Legal Builder Hub.
What is Perplexity’s Computer for Counsel?
Computer for Counsel is Perplexity’s late-June 2026 launch of an agentic platform for lawyers. It connects DeepJudge, Docusign, NetDocuments, Carta, Box, LegalZoom, and Deel via MCP, with Clio and Ironclad integrations planned. Gunderson Dettmer, which has used Perplexity AI firmwide since May 2025, is deepening its use of the product for transactional research and document review.
Which legal tech vendors are most exposed?
Vendors built directly on Claude face the most pressure, because the foundation model now ships connectors and plugins that overlap with their workflow products. Harvey, Relativity, Everlaw, and Thomson Reuters are responding by integrating deeply with Anthropic, a posture LawNext has described as a survival bet as much as a partnership choice.
Will AI replace lawyers?
The vendors closest to large-firm buying patterns argue no. Workstorm COO Nick Stech has said the standard of care in law is too high for “good enough” AI, and his co-founder Raj Fernando has said there is no generic law firm or generic in-house team. The frontier AI companies themselves describe their tools as augmenting legal judgment rather than replacing it.
What is the “last mile” of legal AI?
The last mile is the implementation, change management, governance, security review, and workflow redesign work that turns a foundation model into something a law firm or legal department actually uses. Frontier AI vendors have historically not built this muscle at scale, which is why implementation and consulting partners are emerging as a distinct winner category in the new legal AI stack.
Disclaimer: This article discusses market structure in legal technology and is provided for informational purposes only. It is not legal advice. Decisions about adopting or investing in legal technology should be made in consultation with qualified legal and technology professionals. Figures cited are accurate as of publication.
