Data handling & AI use

How Geminus treats client data and uses AI tools on client work. Last reviewed: August 19, 2026. The engagement contract always governs; this page states our defaults.

Where your data lives

We work inside your infrastructure under your access controls: your cloud accounts, your repos, your warehouse. We do not copy client data onto our machines unless the engagement explicitly requires it, and then only the minimum, encrypted, and deleted at engagement end.

Access

Access is per-person and per-engagement. You approve every engineer before they touch anything. Access is revoked at handoff, and we ask you to confirm the revocation.

AI tools on client work

We use AI coding and analysis tools daily; that is part of how a small senior team delivers. The defaults:

• •  Client data never trains anything. We use commercial API tiers with training disabled, or models running inside your environment.
• •  Code and data stay out of consumer AI tools. No pasting your schema into a free chatbot.
• •  AI-written code ships only after review by the senior engineer accountable for the engagement, and it ships with evals.
• •  If an engagement needs a different arrangement, stricter or looser, we write it into the contract.

Contracts and ownership

US company, US contract. You own all work product: code, documentation, models, and evals. We sign NDAs before first calls when asked.

Questions or a procurement questionnaire: hello@gemin.us.