Data and security
Yours on delivery. Never trained on.
What we do with your prompts, your rubrics and your output, and what we contractually cannot do with them.
Ownership
Assigned in writing, per person, per project.
Everything produced for you is yours on delivery, assigned in writing, including the reasoning traces and the rubric if we wrote it. We keep no licence to resell it, no licence to use it as a sample, and no right to show it to another client.
Every expert signs the same assignment and confidentiality terms before they see a single item, and the signature is on file per person per project rather than as a blanket contractor agreement.
What we never do
No third-party model touches your prompts.
We do not train anything on your material. We do not pass your prompts through a third-party model to speed up drafting, because that is both a confidentiality breach and the exact contamination you are paying to avoid. Where a project allows model assistance, it is named in the contract, the model is named, and every assisted item is flagged in the output.
We do not reuse an item across clients, even where the domain is identical and the item looks generic.
Where the work happens
Your system, your client's, or ours. You choose.
We do not sell a platform and we will not ask you to move your data into one. The work happens wherever suits the project: inside your own tooling, inside your client's, or internally with us, on accounts and workspaces set up for that project alone.
Wherever it runs, access is per person, limited to the project, and revoked the day someone leaves it. Squad leads see their own squad and nothing else. Work that carries regulated or personal data stays in your system or your client's, where your own controls already apply.
Retention
Deleted on project close.
Working copies are deleted when the project closes. Afterwards we keep only the delivery record: a line per batch with a hash, not the content.
When the work runs internally with us and you need it held in a particular region, say so before it starts and we will set it up there.
Security reviews
Send the questionnaire.
If your procurement has a security questionnaire, send it to remotevansh@gmail.com and a person will answer it, rather than a template.
Tell us the domain you are stuck staffing.
One call, and you get a straight answer about whether we can take it. If we cannot, we will say so on the call rather than three batches in.
Prefer to read first? See who runs Xocket Data Labs or every way to reach us.