--- title: "Sovereign Execution™ — sovereignty of the decision, not just the data · KYE Protocol™" description: "Data sovereignty answers where AI runs. Sovereign Execution™ answers who is allowed to act — and proves it. KYE Protocol™ sits above confidential compute and sovereign infrastructure: even when a workload is hardware-attested and isolated, the action it triggers still needs authority, evidence, finality and replay-verifiable proof. KYE™ does not provide data residency or confidential compute; it is the authority and evidence layer on top of whoever does." url: https://kyeprotocol.com/sovereign-execution/ lang: en source: "KYE Protocol" --- > Data sovereignty answers where AI runs. Sovereign Execution™ answers who is allowed to act — and proves it. KYE Protocol™ sits above confidential compute and sovereign infrastructure: even when a workload is hardware-attested and isolated, the action it triggers still needs authority, evidence, finality and replay-verifiable proof. KYE™ does not provide data residency or confidential compute; it is the authority and evidence layer on top of whoever does. # Sovereignty of the decision, not just the data. Data sovereignty answers _where_ AI runs — confidential compute, hardware-attested enclaves, in-region residency. **Sovereign Execution™** answers the harder question your regulator actually asks: _who is allowed to act_ — and proves it. **KYE Protocol™** sits above the sovereign stack. Even when a workload is isolated and attested, the action it triggers still needs authority, evidence, finality and replay-verifiable proof. **Confidential AI protects the computation; KYE Protocol™ governs the consequence.** [Start a governed pilot](https://kyeprotocol.com/poc/) [See how authority is proven](https://kyeprotocol.com/authority-sourcing/) ## Two different questions — and you need both answered If you are a CISO or counsel buying sovereign AI for a UK or EU institution, you are stitching together two layers that are often confused. KYE Protocol™ does not compete with the first; it completes the second. ### Where does the AI run? Data sovereignty Confidential compute, trusted execution environments (TEEs), on-premises or in-region VPC deployment, and data residency. This is what confidential-AI and sovereign-infrastructure vendors solve: the model weights and the prompt never leave a boundary you control, and the hardware attests to it. ### Who is allowed to act — and can you prove it? Sovereign Execution™ Was this agent authorised for this specific action, under this purpose and scope, at the moment it happened? Sovereign Execution™ binds every consequential action to a sealed [Authority Sourcing™](https://kyeprotocol.com/authority-sourcing/) record, an [Evidence Pack™](https://kyeprotocol.com/evidence-pack/), and a Replay-Proof™ verdict anyone can re-check offline. **Honesty boundary for a CISO.** KYE Protocol™ does _not_ provide data residency, confidential compute, or infrastructure sovereignty. It is the authority and evidence layer that runs _on top of_ whoever does. KYE Protocol™ is complementary to confidential AI and sovereign infrastructure — never a replacement for them. ## KYE Protocol™ sits above the sovereign stack The sovereign-AI conversation usually stops at the infrastructure boundary. But a confidential, isolated, hardware-attested workload can still trigger a wire transfer, a denied claim, or a clinical recommendation — and none of that infrastructure tells you whether the _action_ was permitted. Sovereign Execution™ is the layer that does. The bottom-up sovereign stack. Confidential AI hardens everything up to and including the data-sovereignty layer; Sovereign Execution™ governs the consequence above it and emits replay-verifiable evidence. KYE Protocol™ never replaces the layers below — it depends on them. ## What Sovereign Execution™ adds above confidential AI Confidential compute makes the workload trustworthy to run; it says nothing about whether the action was permitted, irreversible-by-design, or defensible afterwards. Sovereign Execution™ closes those four gaps, each as part of the fixed KYE Protocol™ evidence family of 7 signed JSON schemas mapped to the EU AI Act™ Article 12 logging duty, NIST AI RMF and ISO/IEC 42001™. - **Authority — who may act.** Every action is checked at the moment it happens against the acting principal's purpose, scope and limits via [Authority Sourcing™](https://kyeprotocol.com/authority-sourcing/), not against a config row someone could overwrite. - **Evidence — what actually happened.** The inputs, the rules they hit, and the outcome are sealed into an [Evidence Pack™](https://kyeprotocol.com/evidence-pack/) with an Ed25519 context seal, under write-once retention — so a confidential decision is still an auditable one. - **Finality — what cannot be quietly undone.** Authority Finality™ means an irreversible action carries dual-channel sign-off and a revocation is itself a recorded event, so "who turned it off, and when" is never a guess. - **Proof — defensible to a third party.** Replay-Proof™ lets a regulator or auditor re-verify the verdict offline from published keys alone, with no trust in KYE Protocol™ and no access to your confidential enclave. This is not a new engine and not a parallel system. Sovereign Execution™ is the positioning name for what KYE Protocol™ already does — the [Evidence Pack™](https://kyeprotocol.com/evidence-pack/) chain, the [Authority Graph™](https://kyeprotocol.com/knowledge-graph/), write-once audit retention and replay verification — framed for the sovereign-AI buyer who has already chosen where their compute runs. ## Operationalize the regulation — don't route around it For a UK or EU buyer, sovereignty is a means to a compliance end, not the end itself. Sovereign Execution™ is the tool that turns a framework obligation into a per-action control you can show a regulator, mapped across 249 frameworks. ### EU AI Act™ — Article 12 requires automatic event logging over a high-risk system's lifetime. Sovereign Execution™ emits that log per action as signed, replayable evidence, rather than as best-effort application logs. ### For a CISO — when a confidential agent does something it should not have, you replay the exact failed control and revoke mid-flight with Authority Finality™. Isolation alone gives you neither. ### For a CFO — every consequential spend an autonomous agent triggers carries a sealed authority record, so finance can answer "on whose authority was this paid?" as a query, not an investigation. ### For counsel & a regulator — a Replay-Proof™ Evidence Pack™ discharges NIST AI RMF and ISO/IEC 42001™ clause 6 obligations as portable evidence that survives outside the enclave it was produced in. [Start a governed pilot](https://kyeprotocol.com/poc/) [See a sealed decision](https://kyeprotocol.com/evidence-pack/) Canonical KYE™ surfaces referenced on this page: [Authority Graph™](https://kyeprotocol.com/authority-lifecycle/) · [KYE Protocol™](https://kyeprotocol.com/).