Anthropic legal plugin pressures Harvey
Harvey
Anthropic entering legal pushes Harvey further away from being a model company and deeper into being workflow software for law firms. The core legal reasoning is no longer scarce. Harvey already dropped its proprietary legal model after frontier systems from Anthropic, OpenAI, Google, and xAI beat it on BigLaw Bench. What remains defensible is the layer around the model, the review steps, firm specific playbooks, security controls, and integrations into legal research and document systems.
-
Anthropic is attacking the broad, repeatable tasks first. Its legal plugin automates document review, risk flagging, NDA triage, and compliance checks. That overlaps with the top of Harvey’s funnel, but not yet the deeper law firm workflow Harvey sells into large firms and in house teams.
-
Harvey’s product is more concrete than a chat window. Lawyers use it to analyze contracts and filings, pull research through LexisNexis, draft work product, and increasingly build custom agents for their own matters. That turns Harvey into a managed operating layer for legal work, not just a legal prompt box.
-
This follows the same pattern seen in coding. Claude Code pulled foundation models into direct competition with AI coding startups. In legal, the consequence is similar, model labs commoditize the general assistant, while vertical companies either move up into full workflow software or get squeezed by incumbents like Thomson Reuters, LexisNexis, Clio, and Ironclad.
The market is heading toward a split. Foundation labs will own the general purpose legal copilot, while companies like Harvey will win only if they own the systems of work around it, including trusted data access, audit trails, approvals, and firm specific automation. That is also why legal tech is fragmenting into narrower, workflow specific products instead of converging on one universal AI assistant.