For firms and solo practitioners
For matters you can’t paste into a consumer chatbot.
An AI API with a hard spending limit that keeps no prompt or answer content. It is not a chat app: a tool your firm uses connects to it.
One key per matter.
Issue a key for each matter and set its cap. Each month your administrators can download a CSV file for each key, or one for all keys, with one row per day and model: the date, the key’s name, the model, tokens in, tokens out and the cost.
Name keys with a code from your own system, not a client or matter name. Key names are kept in usage records, and those can be requested. Usage records are kept 13 months, or 90 days if your administrator chooses.
KeyCap
- k-0142$150
- k-0157$100
- k-0163$50
Illustrative codes. Not a real account.
First passes, not final work.
-
Depositions and discovery
Summaries organized the way you will use them, with page references to check against the transcript. No PHI: don’t send medical records from a case file.
summarize pp. 14–22; list admissions, inconsistencies and open questions
-
Contracts, first pass
A first read that marks the clauses worth your time. You still read the contract.
flag indemnity, limitation of liability and termination terms that differ from our standard
-
Research memos, drafted
Drafts from material you provide. Check every citation against the source. AI models can invent cases and quotations.
draft a memo outline on this question from the cases I paste below
What a firm should ask us
-
How do we pay?
Practices, firms and companies buy prepaid credit by card, or pay monthly after use, by card or on a company purchase order, up to a monthly cap they set. See pricing.
-
What if you receive a subpoena?
Line Three keeps no prompts or answers, so it cannot produce them; only account, billing and usage records, including key names. We give you at least 7 days’ notice where the law allows, challenge overbroad demands, assert your privilege at your request, and publish a count of requests each January, starting in January 2028. AWS’s records are governed by AWS’s terms. How legal requests are handled.
-
Does this meet my professional obligations?
The judgment is yours, under your jurisdiction’s rules and your bar’s guidance on AI tools. Here are the facts that judgment usually turns on:
- Who receives the material?
- Line Three, then Amazon Bedrock, which runs the model, in the AWS US West (Oregon) Region. No one else on Line Three’s side. Some editors relay your request through the editor maker’s own servers before it reaches the endpoint you set. If yours does, that company is in your request path too. Check your tool’s documentation.
- Is it kept?
- Line Three stores no prompt or answer content. AWS states that, by default, Bedrock does not store model inputs or outputs; the data policy sets out AWS’s terms and links to them.
- Is it used for training?
- Not by Line Three. AWS states that neither AWS nor the model providers use Bedrock inputs or outputs to train models.
- Who can see it?
- No one at Line Three reads it; the gateway decrypts it in memory only to forward it. Afterwards it isn’t kept. Account and usage records are reachable only through administrative accounts with multi-factor authentication, and every access goes to an append-only access log.
-
Can we check what you say?
Ask for the evidence behind any statement on this site and we will answer in writing. Line Three does not offer on-site audits, and holds no SOC 2 report.