How we use AI — and how we don't

How we use it

VettID is developed with substantial help from AI — specifically Claude, made by Anthropic. Specifications, code, infrastructure, and much of this website are produced in collaboration between the founder and AI tooling, with human review, secret scanning, and incremental commits on everything that ships — you can see the work on GitHub. We're telling you this because transparency you have to dig for isn't transparency.

We think this is what responsible AI-assisted engineering looks like: the human is accountable, the AI is a collaborator, and the review process is disciplined.

How we don't

No AI system — ours, Anthropic's, or anyone else's — processes VettID member data. This is not a promise we ask you to trust; it's a consequence of the architecture you can verify:

• Your vault has no inbound access path. Nothing can reach in — not an administrator, not an attacker, and not an AI.
• Your keys are yours. VettID is non-custodial: we can't hand your data to an AI because we don't have your data.
• Delegation, not surrender. When AI agents eventually act for you through VettID, they'll do it under LEASH — scoped delegation with zero secret exposure — not by being handed your credentials.

Where we stand

We think technology — AI included — should serve people, not extract from them. The test of that isn't a mission statement; it's whether the architecture makes extraction impossible. That's the test we build to.

Who reviews all this work? About VettID →