The existing article Sovereign AI is a placement problem sets out the argument: sovereignty is decided one request at a time, by where it runs, who holds the keys and what can be shown afterwards. This post is the practical companion. It lists the choices a person or organisation has when running or buying AI on Tenzro Network, and says what each choice does and does not give you.
Choice 1: where the work runs
Run it yourself. A node is software on hardware you own and host. A model served from your own node on your own machines stays on them. With the --private visibility mode a model is not announced and is served only to you, over a direct or LAN connection. See Compute that stays local.
Choose providers by where they say they are. Every provider declares a coarse region (a country code or a continent-level region, never anything finer) and a jurisdiction. Callers read both from the provider list and can filter on them, and a multi-turn session that asked for continuity can be pinned to one jurisdiction, so that its successor and cache stay inside it.
The limit is stated in the documentation and bears repeating. A region or jurisdiction field is a declaration by the operator, not an audit. It is not a statement of legal compliance in any jurisdiction, and the network does not give one. Where data may lawfully go is for you and your counsel to decide.
Choice 2: who can use a model
A model has one of three visibilities. Public means announced and open to anyone at the provider's price. Gated means announced as access-required, usable only by callers the provider's access policy names, with no pay-per-use path. Private means not announced and not served off the node. These are the access modes of the node. Private is about discovery and serving; it is not a substitute for a security review of the machine that hosts it.
Choice 3: who holds the keys
Account and node keys live in a TPM 2.0, a Secure Enclave or a passkey, never in a file, and a node cannot start without a hardware key. For a machine with no TPM the key lives in another machine's TPM, reached through a key holder that the operator approves with a passkey. The practical test from the sovereignty argument still applies: list every key that can make a node act, and name who holds it. Delegating to a key holder puts that machine on the list.
Choice 4: what you can show afterwards
Metered work yields signed receipts that anyone can verify against chain state. A receipt proves what its fields commit to, such as the model, the serving machine and the input and output hashes. See What a receipt for AI work proves. It does not prove where a machine physically sits.
For work that needs more, a provider can run in a trusted execution environment and attach an attestation. An attestation shows which software ran on which platform at which security level, with the platform vendor's keys as the root of that proof. A TEE narrows who can read data while it is processed. It does not remove trust in the platform vendor, it does not prove the software is bug-free, and a flaw found in a platform weakens the evidence from that platform. See Trusted execution environments.
A short checklist
- Decide which workloads must stay on machines you control, and serve those privately on your own nodes.
- For the rest, filter providers by declared jurisdiction, and read the receipts rather than the marketing.
- Keep identity keys in hardware, and list every key that can act for your node.
- Require attestation only where the threat model needs it, and read what it covers.
- Take legal advice on residency. The network gives none.
Where this fits
Tenzro is an economy for distributed compute and decentralized AI, open to machines of every size, from home nodes to data centers. It works alongside centralized data centers, whose operators are supply and customers of the same rails. Placement is a property of each request, so sovereign work and general work can use the same network. The network is in preview and has not yet carried production use; these controls are what the code provides, not a record of how they have performed in the field.
