Start here
Does any of this sound familiar?
- Your collection includes attorney-client communications, work product, draft advice, or material marked privileged.
- An AI tool may summarize, classify, or evaluate documents outside the matter workspace.
- You need to know what is retained, who sees outputs, and how disclosure is contained.
Privilege risk begins before a document is shared. A prompt, excerpt, embedding, label, or summary may expose substance even when the source file stays put. A confidentiality marking does not itself prove disclosure is protected or authorized.
Separate counsel's use of a tool for a representation from licensing records or training another party's model. Approval for one purpose does not authorize the other.
Not ready to share a single file? You don't have to.
Take the 3-question fit checkThe problem
An inadvertent disclosure rule is not a data-sharing plan
Privilege review depends on the holder, communication purpose, recipients, governing law, and waiver issues. Vendor access, retention, subcontractors, and prompt use affect risk; a “private” label still requires technical and contract review.
Federal Rule of Evidence 502(b) addresses certain inadvertent disclosures, conditioned on reasonable prevention and prompt remediation. It is no blanket guarantee and does not authorize knowingly giving protected documents to an AI provider.
A possible remedy after disclosure cannot replace a prevention and response plan.
The solution
Screen the workflow before the document
Assess categories and configuration first. Counsel must approve a use before submitting a privileged item, excerpt, or derivative.
DataSupply partners only with labs that meet its top 0.01% credibility standard. We help assess whether a qualified buyer may be a fit and negotiate terms that reflect the data's potential value, including exclusivity where relevant. We also help you work through diligence questions about rights, privacy, security, and compliance, then present a high-level inventory of permitted records, not the dataset. Fit is specific to each situation; no buyer or value is guaranteed.
What to inventory before any buyer conversation
- Map privilege and matter authority Identify privilege holder, client instructions, matter limits, protective orders, retention, and processing authority. Separate public material from protected communications and work product.
- Inspect the entire data path Ask where prompts, files, embeddings, logs, outputs, and feedback go; who accesses them; whether providers retain or train on them; and how deletion and incident notices work.
- Approve a bounded test and response record If counsel approves, document task, minimum fields, access, duration, reviewer, and prohibited reuse. Keep configuration, items processed, recipients, and deletion confirmation; define stop-work and remediation steps.
Set the boundaries before discussing access.
Follow matter counsel, client terms, protective orders, and local ethics rules. Require least-privilege access, encryption, no secondary training or retention without approval, incident reporting, logs, and verified deletion. Redaction and synthetic examples do not resolve every rights question.
What could make a permitted example useful?
A synthetic task may permit limited evaluation without protected material; it is not a license to historical legal work. Weigh any efficiency against review, security, and disclosure risks; assume no price or demand.
A practical first step.
Ask a matter partner and security lead to review a vendor data-flow questionnaire using a hypothetical category. Never upload a live privileged file to test the tool.
datasupply.ai can discuss possible fit and buyer questions without receiving your dataset. You decide whether to pursue any introduction. No buyer, license, or payment is guaranteed.
Documented example / what it proves
Rule 502 sets conditions for some inadvertent disclosures
The U.S. Courts' Federal Rules of Evidence state in Rule 502(b) that certain inadvertent disclosures do not waive privilege in federal or state proceedings when the disclosure was inadvertent, reasonable steps prevented it, and the holder promptly rectified it. Read United States Courts, Federal Rules of Evidence, Rule 502.
Those conditions support prevention controls and a response plan. The rule does not make every AI transfer safe, decide whether a record is privileged, or authorize commercial use.
The important limit: Rule 502 is not evidence of a closed data license, a universal no-waiver guarantee, or permission to provide privileged material to a third party.
Where might your own organization stand?
Take the private fit checkQuiz / Your next step
What is the status of the documents proposed for AI use?
If privilege, authority, or the provider's data path is unresolved, pause rather than test with live material.
Your suggested next step
No fee for the initial conversation or introduction. We may be compensated by a buyer if an introduction becomes a partnership. No buyer, license, or payment is guaranteed. Review any proposed deal with your own legal and security advisers.