It depends entirely on where the document actually goes and who can read it while it's there — which most AI-tool marketing doesn't make easy to find out, and which is worth asking directly before you upload anything privileged anywhere.
The question underneath the question
"Is AI safe for privileged documents" isn't really about AI — it's about data handling. A privileged document processed on a vendor's hardware, never logged, never used for model training, with no human reading its content, is a fundamentally different risk than the same document pasted into a consumer chat tool whose terms allow using your input to improve the product.
What to actually check, for any vendor, including this one
Ask plainly: does the document leave your control before it's processed? Is it retained after the job, and for how long? Is it used for anything beyond that one job? Does a human ever read the content? Here, the answers are: processed on hardware we own, retained only for the job's stated window, never used to train anything, and not read by a human except metadata in the rare manual-review case — but a firm's own privilege obligations are the firm's call to make, not a vendor's marketing claim to settle.
Where this breaks down — said plainly
- This is not legal advice about privilege waiver. Whether using any third-party processor waives privilege is a legal question for the firm's own counsel, not something a vendor can answer for you.
- Pattern-based tools don't make privilege calls. Automated extraction and redaction find known patterns; they don't exercise legal judgment about what's privileged.
Related guides
- What to check before using AI on HIPAA data
- Local AI vs. cloud AI for documents
- Remove hidden metadata from a Word document
Where this fits on the site: this is part of how every job runs here, across all services. See Services & Pricing for what's actually sold, or go straight to Choose your job.