Our Take
Three big tech companies launching legal offerings in 12 months is not about legal expertise; it's about controlling where lawyers spend their attention and which LLM they depend on.
Why it matters
Law firms and in-house teams now face a structural choice between entrenched legal tech companies (which can mix LLMs) and big tech vendors (which lock you into a single model ecosystem). The outcome will reshape how legal tech companies position themselves for the next three years.
Do this week
CTO or head of legal tech: audit your current contract review and document automation workflows this month to map exactly which tasks have real switching costs if you move to a big tech platform.
OpenAI recruits legal talent to launch vertical product
OpenAI is planning a legal-specific offering under the working title "Codex for Legal," following a strategy already deployed by Anthropic and Microsoft. The company has approached at least one executive from a contracts-focused legal tech company and is considering additional senior hires from the legal tech sector (per Artificial Lawyer sources).
The product would fit into OpenAI's broader "Codex for..." framework, a platform launched in April that extends beyond coding into domain-specific verticals like graphic design and sales. OpenAI's blog post that month noted that Codex can now interact with desktop applications through computer use, click and type with its own cursor, and work with over 90 plugins that combine skills, app integrations, and MCP servers.
Like Anthropic's Claude for Legal (which expanded to 12 plugins and multiple MCP connectors last week), OpenAI's offering would likely ship as a suite of enterprise plugins tied to legal tech software. Integration with ChatGPT remains undefined. So does ease of use: Claude for Legal, despite operating in a no-code environment, requires technical support from most law firms and in-house teams, though large firms typically have engineering staff on hand.
This move follows OpenAI's announcement of the "OpenAI Deployment Company," a forward-deployed engineer (FDE) program designed to help enterprise customers operationalize LLM capabilities. That infrastructure could accelerate legal adoption.
The real battle is over workflow lock-in, not legal accuracy
Three major technology vendors now competing for the center of the legal workflow represents a fundamental shift in how law firms acquire AI capability. The question is no longer whether AI can review contracts or summarize documents; the question is where lawyers will "live" when they do that work.
Microsoft's Legal Agent integrates with Word, a lawyer's default environment. Anthropic's Claude for Legal offers Claude for Word and Cowork, positioning itself alongside daily work. OpenAI's broad individual and enterprise adoption gives it a distribution advantage that neither competitor fully matches.
Incumbent legal tech companies face a specific pressure: they can mix and match LLMs as models evolve, but they cannot match the pricing power, distribution, or brand pull of big tech. At the commoditized end of legal work—basic document review, contract analysis—vendors with no lock-in dependency will face margin pressure. Companies with sticky workflows, customer relationships, and proprietary legal domain knowledge retain an edge, but only if firms choose to keep them.
For law firm CTOs and in-house heads of tech, this is not just three new options. It is a decision point: do you stay modular and risk fragmentation, or do you pick one vendor and accept dependency on their LLM roadmap?
Map your switching costs now
Before you pilot any of these offerings, inventory which legal workflows would actually move if you adopted a big tech platform. Not every task is equally portable. Contract review against a specific clause taxonomy, integration with your matter management system, or custom training on your firm's legal language—these have switching costs that matter.
Talk to your current legal tech vendors about their LLM flexibility and roadmap. Ask whether they plan to support multiple models or tie themselves to one. If you have the budget and no strong current vendor dependency, running a parallel test of both Claude for Legal and Codex for Legal (once available) will be cheaper than making a wrong long-term bet.
Do not assume that "easier to use" means "better for your firm." Ease of use and integration depth are different things. Cowork and Word plugins are easy; they are not always the best fit for custom legal workflows.