AI for law firms

Is ChatGPT safe for confidential client information?

There is no blanket safety guarantee. Before using client material, your firm must approve the product configuration, access, retention and data terms against its confidentiality duties, client instructions and data-use policy. Start with anonymised or synthetic examples. A firm-controlled business workspace can support access controls, but does not itself establish that live matter data is permitted. Lawyers remain responsible for deciding which material may be used and checking every output.

Where the risk actually comes from

Confidentiality risks include product settings and data terms as well as staff behaviour and unclear permissions. A reviewed setup answers three questions before anyone starts: which approved account, who can see what, and which material the firm and client permit. Access controls reduce some risks; they do not establish legal compliance.

Swipe to read every column, or focus the table and use the arrow keys.

QuestionUnsafeControls to review
Which account?Personal or free accounts, chosen by each lawyerOne business workspace owned and administered by the firm
Who can see it?Anyone with a shared loginNamed users; projects and Skills shared only with the matter team
Which material?Whatever is to handA written data-use policy by material type, agreed before use
Which terms apply?UnreadThe plan’s current data terms checked by the firm before rollout and reviewed when they change
Who checks outputs?NobodyThe responsible lawyer reviews every draft before it leaves the firm

A data-use policy, in brief

Allowed in the firm workspace:
  precedents, clause library, style guide, anonymised past work
  live matter documents only where firm policy, client instructions and product terms permit; matter-team access alone is not approval
Remove or anonymise first:
  identity documents, bank details, health information
Never in any AI tool:
  material a client has said must not be processed this way
  documents under a court or regulator restriction on sharing
Always:
  firm workspace only; no personal accounts or browser plug-ins
  lawyer review before anything leaves the firm

Sample policy for illustration. Each firm sets its own rules with its partners and compliance lead.

Guardrails

  • Partners approve the data-use policy, product configuration and access, and check client permissions before any live matter material is used.
  • Lawyers review and sign off every draft, summary and client update.
  • AI never sends anything to clients, courts or other parties on its own.
  • Clients who ask are told how AI is used on their matter, in line with your engagement terms.
  • Your document management system stays the record; AI works beside it. See our ChatGPT governance guide.

Where to start

Many firms begin with work that needs no client data at all, such as a Skill built on precedents and the clause library, for AED 97, and agree the policy before moving to live matters. Firm-wide setups with access rules are fixed-price packages from AED 3,500; see the pricing page.

Common questions

Is ChatGPT safe for confidential client information?

There is no blanket guarantee. The firm must approve the specific product configuration and data terms against firm policy and client instructions before using live matter material. Start with anonymised or synthetic examples.

Can lawyers use their personal ChatGPT accounts for client work?

Use only firm-approved products and data handling. A business workspace or matter-team access does not by itself permit client material; firm policy, client instructions and product terms must allow it.

Should we tell clients we use AI?

Many firms do, in their engagement terms or when asked. The lawyer remains responsible for the work and its confidentiality either way.

Can we start without using client data?

Yes. Skills built on your precedents, clause library and style guide need no client material, so firms often start there while the policy is agreed.