ChatGPT Business costs a small law firm $20 per Standard seat a month on annual billing ($25 monthly), with a two-seat minimum, so the floor is $40 a month. OpenAI does not train on workspace content by default. The settings that matter are who can read chats, retention, project sharing and connected apps.
Business is self-serve, and OpenAI's own help page says that Zero Data Retention and other sales-led terms need a contracted plan instead. Whether your firm may put client material into any AI tool depends on your regulator's guidance, your engagement letters and your insurer. Treat the settings below as the technical half of that decision, not the whole of it.
What a small firm pays: three firm sizes in dollars
OpenAI sells two seat types inside Business. Standard seats are $20 a user a month if billed annually and $25 if billed monthly. Premium seats are $100 annually or $125 monthly and remove the five-hour usage limit. You can mix the two in one workspace, and the workspace must hold at least two seats in total (two of either type, or one of each). OpenAI's pricing page shows teams of 2 to 200.
The table shows what the seats cost for three illustrative firms. The six-person and twelve-person firms give a partner-level seat to the heaviest user, since Premium is for people who hit limits, not for everyone.
| Firm | Seats | Monthly, annual billing | Yearly total | Monthly billing instead |
|---|---|---|---|---|
| Three people (two lawyers, one paralegal) | 3 Standard | $60 | $720 | $75 a month, $900 a year |
| Six people | 5 Standard, 1 Premium | $200 | $2,400 | $250 a month, $3,000 a year |
| Twelve people | 10 Standard, 2 Premium | $400 | $4,800 | $500 a month, $6,000 a year |
The seat bill is the smaller cost. Add the time to set the workspace up and to write and teach a policy. Say the six-person firm spends three hours on the admin settings, three hours drafting a one-page rule, and an hour per person on a training session: twelve hours. At an internal cost of $80 an hour that is $960 once. Spread over a year with the $2,400 of seats, the first year costs about $3,360, or $560 a person. From year two the one-off drops away, though a refresher session after any change to the plan is worth keeping.
Three budgets
- Cheap: two Standard seats on annual billing, $40 a month, with the owner doing the settings in an afternoon. Good for a firm that wants to pilot on non-client work first.
- Middle: the six-person line above, $200 a month, plus a written policy and a half-day pilot.
- High: twelve seats with two Premium, $400 a month, plus a compliance review of the engagement letter. At this size ask OpenAI about Enterprise terms, because the controls in the later section start to matter.
As a sense of proportion, a Standard seat at $20 a month is eight minutes of a lawyer's time valued at $150 an hour. The cost question is rarely whether the seat pays back. The real cost sits in review time and in a mistake, which the later sections address. The tutorial on when a small firm should move from Business to Enterprise covers the point at which that changes.
The Complete Excel Course with AI Integration
From formulas and pivot tables to Power Query, DAX, VBA and AI with Copilot, ChatGPT and Claude — hands-on from the first lesson.
By Sagnik Bhattacharya and Virginia Thorn, Primetime Emmy Award winner
Buy nowThe two-seat minimum, the seat ceiling and what removing a person does
Seats and members are separate things in Business. The owner buys paid capacity, then assigns it. Three practical effects follow.
- Removing someone does not stop the bill. OpenAI's help page says removing a member ends their access immediately but leaves the seat in the billable count. The owner must remove the seat itself, in Workspace settings under Members, then Users and Manage seats, to stop paying from the next billing cycle.
- Cutting seats is slow. OpenAI's billing help says seat reductions take effect at the start of the next billing cycle, unused seats are generally non-refundable, and on annual plans added seats stay committed until the end of the annual term. A seat bought for a locum or a work-experience student on annual billing is paid for until renewal, so reassign a vacated seat to them instead, or stay on monthly billing until headcount settles.
- Newer workspaces have a ceiling. Pricing says Business is for 2 to 200 people, and workspaces created after 24 August 2026 are limited to that range. A firm that expects to pass 200 users is already in Enterprise territory.
ChatGPT Business used to be called Team. OpenAI renamed it on 29 August 2025 without changing the plan, so older guides and search results about ChatGPT Team describe the same product at older prices.
Privacy settings a law firm owner should check before the first prompt
This table lists the settings and commitments, what OpenAI says, and what to do about each. Work through it as the owner, with the workspace open in another tab.
| Area | What OpenAI says | What to do |
|---|---|---|
| Training | Business workspace content is not used to train models by default, and the same goes for data reached through connected apps | Read the Business terms once and note the date. Do not rely on a staff member's personal account, which has different defaults |
| Retention | The Business FAQ on OpenAI's privacy page says workspace admins can control how long data is retained, and deleted or unsaved conversations leave OpenAI's systems within 30 days unless the law or protection of the service requires longer. The same page's summary lists retention control only for Enterprise, Edu and Healthcare, and the pricing comparison puts custom retention policies under Enterprise | Open your own admin settings and see what the control actually offers. Write the setting you chose and its date in the firm's IT record |
| Sign-in | SAML SSO, MFA, domain verification and an admin console are included | Turn on MFA for everyone and verify the firm's domain, so staff cannot create a rival workspace on the firm's email addresses |
| Memory | Shared projects automatically use project-only memory and cannot see a member's other memories | For matter work, use projects with project-only memory, and decide whether staff should switch memory off for their other chats (owners and admins can manage the workspace's supported memory settings) |
| Connected apps | Admins choose which apps are enabled, each user signs in to a connected app themselves, and ChatGPT respects that app's permissions | Enable only the document store the firm already controls. Leave email and calendar apps off until you have decided the policy |
| Project links | Projects can be shared with named people, workspace groups or a workspace link, which gives every member chat access by default | Invite individuals on the matter only. Never use the workspace link for a client matter |
| Contract | OpenAI can execute a data processing addendum for Business customers | If your clients or insurer ask for one, request it before the pilot, not after |
The retention row is the one to be careful with. OpenAI's own pages describe retention control for Business in two different ways, so what your firm gets depends on what the admin console shows. Do not copy a retention promise into a client letter until you have seen the setting yourself. For the full list of admin switches, the admin console checklist for ChatGPT Business goes through each page in turn.
Two official pages are worth reading in full before you decide: OpenAI's business pricing and plan comparison and its enterprise privacy page.
Business or two personal Plus plans: same money, more control
A two-lawyer firm often starts by asking whether each lawyer should just buy ChatGPT Plus. Plus is $20 a month billed monthly, so two Plus plans cost $40, the same as two Business seats on annual billing and $10 less than two on monthly billing. The price is almost identical, and the difference is everything around it.
- Training default: OpenAI's data controls page says that on personal workspaces with Free, Plus or Pro, data sharing is on by default, and each user has to switch off the Improve the model for everyone setting. In Business that is reversed: no training by default.
- Ownership: a personal account belongs to the person. If a lawyer leaves, the chats go with them and the firm cannot see or delete them. A Business workspace belongs to the firm, and the owner can remove the member and keep the workspace.
- Shared work: Plus can share a project with a small number of collaborators, but there is no admin console, no domain verification and no way to apply one setting to everyone.
- Evidence: if a client or insurer asks how the firm controls AI use, "each lawyer has a personal subscription" is a weaker answer than "we run a workspace with MFA, named admins and a written policy".
So the two Business seats cost a few dollars more at worst, and for the same money they buy the controls the rest of this tutorial depends on. If you are a sole practitioner, the two-seat minimum is the sticking point, and the further reads below cover the options when you only need one login.
Who inside your firm can read whose chats
Lawyers worry about OpenAI reading their chats, and they should also ask who in the firm can. OpenAI's privacy page for Business says users see their own conversations, and workspace admins can view, access, export and delete end-user conversations in the workspace. OpenAI limits its own staff's access to conversations to engineering support, abuse investigation and legal compliance, plus contractors bound by confidentiality.
The internal point matters for a firm with an information barrier, for example between two matters on opposite sides of a dispute, or a partner reviewing a colleague's performance. An admin is usually the office manager or a partner, and that person can open a chat a fee earner assumed was private. Three simple rules reduce the risk:
- Keep the admin role to two named people, one of them a partner.
- Tell staff plainly in the policy that admins can see workspace chats, so nobody uses the workspace for personal or HR matters.
- Where a barrier applies, use a separate project with named members and do not discuss the matter anywhere else in the workspace.
This is a firm-governance decision, so take it to whoever handles your professional-conduct questions rather than settling it in the settings page alone.
What Business lacks: audit exports, zero retention, residency
OpenAI's feature comparison is blunt about what sits on the Enterprise side. For Business it shows these as not included: the Compliance API logs platform, SCIM, role-based access controls, IP allowlisting, Enterprise Key Management, and data residency in the listed regions. Business does include SOC 2 Type 2 compliance, a basic analytics dashboard and shared projects.
For a law firm those gaps translate into four plain questions.
- Can we archive prompts and answers into the file? Without the Compliance API, you rely on admin export from the console. Test that with a dummy conversation, because a firm that needs a durable record of AI use on a matter may find it incomplete.
- Do we need Zero Data Retention? OpenAI's help page says that if you need it, or other sales-led terms such as invoice billing, you use a contracted offering rather than self-serve Business.
- Must the data stay in a particular region? OpenAI's pricing comparison, its business-data page and its data residency help article all offer a storage region only to eligible Enterprise, Edu, Healthcare and API customers, not to Business. If a client contract names a region, treat Business as unable to meet it and ask OpenAI in writing about Enterprise.
- What if the output is wrong or infringing? OpenAI's Service Terms give the output indemnity, which covers third-party intellectual-property claims over output, to API customers and to Enterprise customers, a term the same terms define as Enterprise, Edu and Healthcare. Business is not on that list, so do not assume the plan protects the firm from a copyright claim over generated text, and read the current terms before relying on any indemnity.
If two or more of those matter, the answer is usually Enterprise or a legal-specific product rather than Business. The tutorial comparing Claude and ChatGPT for a law firm looks at the same question from the vendor side.
Shared Projects as matter templates: a lease-review project filled in
A Project keeps instructions, files and chats together, and Business projects hold up to 40 files and up to 100 collaborators. Members get either chat access, which lets them use the project but not change it, or edit access, which lets them change instructions and files and invite others. Deleting a file from a shared project deletes it for everyone. Together that makes a project a good home for a reusable matter template, provided the instructions say what the assistant must not do.
Here is a filled-in set of project instructions for a commercial lease review, written to be pasted into the project's settings. It is illustrative, so adapt it and have the supervising lawyer sign it off.
Project: Commercial lease review (template, no client details)
Role: you help a lawyer review a draft lease. You do not give legal advice.
Rules:
1. Use only the documents in this project. If something is not in them, say "not in the documents".
2. Quote clause numbers and the exact words for every point you raise.
3. Output a table: clause | what it says | possible issue | question for the lawyer.
4. Mark any point that depends on a statute or case law as NEEDS LAWYER. Do not state the law.
5. Never include client names, addresses or reference numbers in your replies; use "the tenant" and "the landlord".
6. End with a list of anything you could not check.A reviewer ran a draft lease through a project like this. An illustrative extract of what came back:
Clause 14.2 | Tenant may assign the lease freely | Unusual: free assignment | Confirm this is intended
Clause 9.1 | Rent review every 3 years, upward only | Common | Check the review formula in Schedule 2
Clause 22 | Break clause on 6 months' notice | Check whether it needs conditions | NEEDS LAWYERThe fix is on line one. In the draft, clause 14.2 actually allows assignment only with the landlord's written consent, which the assistant had paraphrased wrongly. Rule 2 is what made the error catchable, because the quoted words did not match the clause. Add a line to the project telling reviewers to check every quoted clause against the document before relying on the table. The same discipline appears in the prompts for lawyers with confidentiality red lines.
Three data tiers for a law firm's prompts
The settings only help if staff know what goes where. A three-tier rule is easier to follow than a long policy.
| Tier | What it covers | Where it may go |
|---|---|---|
| Green | Public law explained in general terms, templates with no matter details, the firm's own marketing, internal admin | Business workspace, no special steps |
| Amber | Matter facts with names, places, dates and numbers removed or replaced | Business workspace, inside a named project, with the supervising lawyer's approval |
| Red | Privileged advice, unredacted documents, anything under an information barrier, anything a client has told you not to share | Not in any general AI tool unless the engagement letter and the client's consent cover it |
Here is an amber-tier before and after. The raw note would be red.
Before: Mr [surname] of [street address] was dismissed on 3 March after
an incident with a customer, ref MAT-2291. Draft a letter to the other
side's representative setting out our position.
After: An employee was dismissed after an incident with a customer.
Draft an outline of a letter to the other side's representative setting
out the client's position on the dismissal process. Use [Employee],
[Client] and [Date] as placeholders.The rewritten version keeps the task and drops everything that identifies the person or matter. The lawyer puts the details back after the draft returns. That puts a small cost on every use, which is the point: it is slower than pasting the file, and quicker than a breach conversation.
Engagement-letter and staff-policy wording to adapt
Two short pieces of wording close the loop. They are illustrations only, for your own lawyer or insurer to review and edit.
For the staff policy: "Staff may use the firm's ChatGPT Business workspace for green and amber material only. Red material must not be entered. Every output is a draft that a qualified person checks against the source before it is used. Workspace administrators can see conversations in the workspace, so do not use it for personal matters." That is four sentences, and it answers the questions staff ask most.
For client-facing wording, some firms add a sentence to their terms of business: "We may use an AI assistant within a secured business workspace to help prepare drafts. We do not enter your identifying information without your agreement, and all work is reviewed by a lawyer." Whether that wording is enough, or whether you need explicit consent for a given matter, is a professional-conduct call for your regulator's guidance and your own advisers. The tutorial on using ChatGPT without breaching confidentiality sets out how firms usually approach that conversation, and the list of legal tasks to keep away from AI is a useful companion for the red tier.
A four-week pilot with checks at each stage
- Week 1, two seats: buy two Standard seats on monthly billing ($50 total) so the commitment is small. Complete the settings table. Run only green tasks, such as rewriting a website page or a client newsletter.
- Week 2, canary check: in a test project with two members, enter a made-up phrase from one account. Confirm what the other account and the admin can see, delete the chat, and note the date. Remove the second member and confirm access ends, then check whether the billable seat count changed.
- Week 3, amber tasks: take three past matters, anonymise them using the before-and-after pattern, and run the lease-style project template on each. Log time saved and errors caught.
- Week 4, decide: compare the log with the seat cost, show the partners the admin visibility point, and decide whether to move to annual billing, add seats, or stop.
Keep the log as the evidence for your decision, whatever it says. If the pilot shows that your matters are mostly red tier, the sensible outcome is a narrower use, such as marketing, internal knowledge and drafting from templates, or a different product, and not a looser policy.
Questions law firms ask about ChatGPT Business
Can a one-lawyer firm buy ChatGPT Business?
Only with two paid seats, because the plan needs at least two Standard or Premium seats in total. That is $40 a month on annual billing or $50 monthly. A sole practitioner either buys the second seat, uses it for a trusted assistant, or stays on an individual plan, which lacks the workspace controls and business terms.
Does ChatGPT Business train on our client documents?
OpenAI states that it does not train on ChatGPT Business workspace data by default, and its Business terms and enterprise privacy page repeat the commitment. That covers content in chats and projects, and data reached through connected apps. It does not replace your own duty of confidentiality, so the firm's policy still governs what staff may enter.
Can the firm keep a record of every AI conversation?
Not through the Compliance API, which OpenAI's pricing comparison lists for Enterprise only. Business admins can view and export conversations in the workspace according to OpenAI's privacy page, but you should test how complete that is before relying on it for record-keeping, and ask OpenAI about Enterprise if you need an audit trail.
Should we use custom GPTs for matter templates?
Use shared Projects instead. OpenAI is retiring custom GPTs, and existing ones are scheduled to stop running on 11 December 2026, while Projects hold instructions, files and chats together and are supported on Business with chat or edit access. A project is also easier to review, because the instructions sit in one visible place.
Further reads
- ChatGPT or a Legal AI Tool: Which Should a Small Firm Use? — When a legal-specific tool beats a general assistant.
- Can Lawyers Use AI Note Takers in Client Meetings? — Recording and summarising client meetings safely.
- AI Implementation Plan for a Small Law Firm: The First 90 Days — A 90-day plan once the workspace is configured.
- Claude v ChatGPT for Lawyers: A Same-Task Test on Your Own Documents — Compare assistants on your own documents.
- AI Setup for a Two-Lawyer Estate Planning Firm: The First Month — A week-by-week first month for two estate planning lawyers: the policy, the account, one workflow, the client notice and the review, with hours and costs.
- E-Discovery Software for Small Law Firms: Costs, Limits and AI Review — Six e-discovery pricing models, one matter costed from your own data, and a sampling test for AI review, written for a firm with a handful of lawyers.
- AI for Immigration Lawyers: Docketwise or ChatGPT for Forms? — Run one immigration form through Docketwise and ChatGPT: what each catches, the cost per filing, the privacy settings and the steps that stay manual.
- Free ChatGPT for Lawyers: Where the Free Plan Stops Being Safe — Free ChatGPT suits generic legal drafting, not client facts. What the Free plan keeps, how to anonymise a matter, and when a business plan becomes necessary.
- Lexis+ With Protégé Pricing: Online Plans for Firms of One to Three — What a one-to-three-lawyer firm can learn about Lexis+ with Protégé pricing online, plus the renewal terms and extra charges to check.
- Personal Injury Lawyer Chatbot: Intake Questions, Limits and Cost — An intake script, disclaimer wording, handoff triggers and monthly cost bands from named vendors for a personal injury firm's website chatbot.
- AI Discovery Review for Defence Firms: Index, Search and Cross-Check — Index discovery, map each statement to the disputed issues, build a contradiction table with page cites, then verify every cite by hand before filing.
- AI Legal Bill Review: Fix Time Narratives Before Invoices Go Out — A pre-bill routine for small firms: tidy narratives, screen for privileged detail, test client billing guidelines, then a partner signs off exceptions only.
- AI Legal Brief Writing: A Draft and Cite-Check Routine — A routine for small litigation firms: build an authority table, draft only from it, cite-check each source line by line, and meet court AI-disclosure rules.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: OpenAI's ChatGPT Business pricing and feature comparison page, ChatGPT Business and Enterprise privacy pages, and OpenAI's help articles on Business seats, members and Projects, read on 1 November 2026.