Google's Gemini Enterprise for Legal, Explained

On 25 August, Google Cloud walked into the room where law firms bill by the six-minute unit and put a product on the table aimed straight at it. Gemini Enterprise for Legal is a version of Google’s enterprise AI platform tuned for law firms and corporate legal teams, and it launched in preview with names that make managing partners sit up: Cleary Gottlieb, Freshfields, Weil, and Williams and Connolly are among the first firms using it.

This is Google’s first industry-specific packaging of Gemini Enterprise, and the legal edition arrived alongside a financial services one, with healthcare and life sciences versions flagged as coming. The pitch goes beyond another chatbot bolted onto Word: a set of pre-built agents and ‘skills’ that try to carry a legal task from start to finish inside the firm’s own systems.

Table of Contents

What it actually does

The platform comes with legal-specific skills that walk agents through recognisable jobs rather than generic prompts:

Drafting legal briefs and verifying citations against primary authority.
Contract review, due diligence and contract lifecycle management.
Regulatory horizon scanning, so a team gets told when the rules shift.
Fulfilling data subject access requests, the DSAR grind that eats junior time.

The part firms will care about most is plumbing. Gemini Enterprise for Legal connects into the document management, e-discovery and legal research systems firms already run, including tie-ins to the Thomson Reuters legal-tech world. Through those secure connectors it inherits a firm’s existing ‘ethical walls’, so the AI is meant to respect who is allowed to see which matter. Google says research outputs are grounded in primary legal authority rather than model training data, and that a firm’s files, playbooks and negotiated positions stay inside its own perimeter.

Why Google is doing this now

General-purpose AI has hit a ceiling in regulated professions. In law it is not enough to produce a fluent answer. The system has to respect confidentiality, access rights and internal policy, and it has to show its working against real sources. Google is betting the next stage of enterprise AI is vertical: infrastructure built for a specific profession where context and control matter as much as raw model quality.

It is also a direct move on turf that Harvey, Thomson Reuters CoCounsel and a pile of legal-AI startups already claim. Google turning up with distribution, a preview roster of elite firms and the Gemini brand is the kind of competition that reshapes a market’s pricing overnight.

The catch, and it is a real one

Citation verification is front and centre of the pitch, and there is heavy irony in that. The reason legal AI has to advertise citation-checking at all is that earlier tools kept inventing cases, and real lawyers have been sanctioned by real judges for filing hallucinated citations. Selling ‘we check the citations’ is an admission of how badly the underlying problem still bites.

Then there is the word ‘preview’. This is not generally available, and a roster of four marquee firms tells you very little about how the thing behaves in a 40-lawyer shop with no dedicated innovation team. ‘Grounded in primary authority’ is a claim to test, not a guarantee to trust, because grounding reduces hallucination without deleting it. And handing your matter management, playbooks and negotiated positions to a single vendor’s connectors is a lock-in decision that outlasts whichever partner signed off on the pilot.

For firms whose entire product is defensible judgement, the sensible posture is to treat these agents as a very fast paralegal whose work you still read line by line. Useful, quick, and absolutely not the person who signs the filing.

Did you know: the first widely cited case of a lawyer punished for AI-invented citations came in 2023, when a New York court fined two attorneys 5,000 dollars after they filed a brief stuffed with fake cases produced by ChatGPT.

Sources

The free stack. One email a week: the AI tools and moves that actually matter, hype filtered out. Subscribe free →