Customer lists, financials, and internal processes are not harmless inputs. Here is what happens to them, who can touch them, and where you keep control.
Your proprietary data is never used to train public models. It is used only to do the work you hired me to do.
I only ask for the data a specific workflow needs. If a field, export, or system is not necessary, I do not need access to it.
These are the controls I use. If an engagement needs something stricter, we agree on it before data moves.
Minimize first. Protect what remains. I request the narrowest useful slice of data. An NDA is available on request.
Retention and deletion. You can ask for your working data to be deleted at any time, and by default it is deleted within 30 days of an engagement closing. Legal and accounting records may still need to be kept.
Encryption by default. Data is encrypted in transit and at rest on the systems used for the work.
Least-privilege access. Access is limited to what the work requires, protected with MFA, and not shared. Your data stays separate from other clients' data.
Sub-processors are disclosed. The LLM providers and automation platforms a workflow needs may process the limited data it requires, under business and enterprise terms that do not train on your data. The specific providers are named in each engagement's scope.
You keep the judgment. Nothing customer-facing sends, and no production change goes live, without your review. Approval gates are part of the build, not an afterthought.
I do not claim certifications this practice does not hold. I name the controls that are actually in place, put handling requirements in writing, and tell you which third parties the workflow depends on.
If your data has regulatory, contractual, residency, or internal-policy restrictions, tell me before the engagement. I will match the tooling to those limits or keep that data out of AI processing.
Ask it in writing. I will answer in writing before you share anything sensitive.