AI Safety & Training
Two halves of one problem. The safety half reviews every AI tool your firm touches, formal or informal, against the vendor's data-handling terms and your actual settings, and answers two questions: is client data being handled safely, and is the output reliable enough to act on. The training half teaches your attorneys and staff to use the tools on your real work: prompting for legal tasks, verifying output before it enters work product, and what supervision looks like when AI is in the workflow. It is a configuration review and hands-on training, not a security certification and not an ethics course.
What is included
- An inventory of the AI tools in use, including informal ones like consumer ChatGPT, Copilot, or Gemini, and a short interview about how they are actually used.
- A review of each vendor's data-handling terms: training, retention, opt-out status, and whether a SOC 2 report exists.
- A review of your current settings for each tool: whether inputs can reach vendor training, whether access is role-appropriate, whether logging is on.
- A red, yellow, green risk register with a plain-English explanation of each issue and the exact setting to change.
- One or two live training sessions on your tools and your real task types, using placeholder data, never live client information.
- A custom prompt library for your common tasks, a one-page verification checklist, a session on supervision, and a 30-day follow-up call.
How it runs
- You send a tool inventory using a provided template; a brief interview confirms real usage patterns.
- JurisLabs reviews the vendors' terms and your settings, then delivers the risk register and walks you through the fixes.
- Training materials get tuned to your tools and practice areas, and the live sessions run.
- Materials delivered, with a 30-day follow-up call for the questions that surface in real use.
What you get
A written risk register for each tool with a prioritized remediation checklist, the settings corrected, a custom prompt library, a one-page verification checklist, and live training sessions with a 30-day follow-up call.
About 2 to 4 weeks from the tool inventory to the last session and materials.
A good fit if
- You adopted AI tools, formally or informally, and never rigorously reviewed what they do with client data.
- Your attorneys or staff use AI on client work without training or a supervision structure.
- You want documentation you can hand to the next new hire.
Not the right fit if
- You have not adopted any AI tools yet. Start with an AI & Automation Build; its scoping phase picks the right tools first.
- You want a formal compliance certification, a CLE ethics seminar, or a written ethics opinion.
Examples
- An attorney drafts client letters in free-tier ChatGPT, where inputs default to training. The register flags it red and gives the exact setting and tier to switch to.
- A firm uses Microsoft Copilot but never enabled the data-protection settings or scoped document access. The register lists the specific settings to turn on, and the training covers what still needs human review.
- A solo immigration attorney gets a prompt library for cover and hardship letters and a checklist for verifying citations and client-specific facts before anything enters the file.
What is not included
- A penetration test or formal security audit of any vendor's infrastructure.
- Any opinion on whether your AI use complies with your bar's professional-responsibility rules.
- A review of your wider IT infrastructure beyond where it touches AI data flows.
- Ongoing or retainer training. This is a bounded engagement.
Talk it through
Book a free call. You will leave knowing whether this is the right starting point for your firm, even if you never hire JurisLabs.