Regulated firms can't outsource the decision and the defense of the decision. Non-custodial audit is the architecture that keeps both in your hands — your data, your model, your audit manifest, your defense.
Custody is the question regulated firms skip. If the vendor holds the data, the model, and the audit trail, the firm is defending decisions it didn't make, on infrastructure it doesn't control. That's not a defensible position in front of a regulator.
The Law of Conduct: Own the Decision and the Defense
A regulated decision needs an owner who can defend it. The conduct rule: the firm owns the data, the keys, and the audit manifest — the vendor owns none of those. The defense is built from the manifest, and the manifest stays with the firm.
- Your data, in your environment.
- Your keys — BYO LLM keys, not the vendor's.
- Your audit manifest, attached to every output.
The Power Dynamic: Custody Is Leverage
Custody is leverage — whoever holds the data and the audit trail holds the firm. Non-custodial architecture keeps the leverage with the firm, not the vendor. The regulator's question is answered from the firm's own records, not a vendor's export.
What the Audit Manifest Carries
Every Deino output carries an audit manifest: the model, the data, the version, the lineage, the decision. Six months later, the firm can reconstruct and defend the decision from the manifest — without the vendor, without re-running the model, without hoping the records still exist.
How FACTA Frames It
Deino is built for the compliance officer who has to defend every AI decision that ships — audit manifests on every output, invariants in the Rust workspace, BYO LLM keys. The firm owns the decision and the defense; the infrastructure supports both.
Conclusion
Non-custodial audit keeps the decision and the defense in the firm's hands. Own the data, the keys, and the manifest — and the regulator's question is answered from your records, not a vendor's.
About FACTA
FACTA helps startups and growth-stage teams turn AI into production systems that keep running — not demos that impress once.
We design the architecture around the parts that actually break under real usage: tooling you own, credentials you control, failover, cost controls, observability. The boring infrastructure that keeps a system alive after launch.
Led by Matías Baglieri and Carolina Fogliato, we focus on one thing:
AI leadership that builds. Not just advises.
Ask us who holds your AI audit trail today.
We'll tell you whether you can defend a decision from it without your vendor. See forecasting infrastructure for the substrate.
Explore AI Automation