PROJECTS APPROACH FOUNDER INSIGHTS CONTACT
Verifiable compute

Aethelred

Make machine outputs provable, governable, and portable.

Project source ↗
DomainVerifiable compute
FocusAttested AI, proof, policy, and deterministic settlement
Research stageProtocol in Testing
Research thesis

A machine output is not trustworthy merely because the model that produced it is capable.

Aethelred is designing a sovereign Layer-1 network for regulated AI and verifiable compute. Its purpose is to bind an output to evidence of what ran, where it ran, which policy applied, and how the result was verified.

The proposed unit of trust is a Digital Seal: a portable record created only after confidential execution, dual-evidence verification, and deterministic settlement. The architecture is public, but the project itself describes it as design-stage and governance-dependent.

06Execution stages

Commit through Digital Seal

04Network layers

Consensus, execution, verification, settlement

02Evidence paths

TEE attestation plus zero-knowledge proof

Fail closedVerification rule

No settlement when either signal fails

The trust deficit

Regulated systems need evidence, not model confidence.

A model can return a plausible answer while leaving no durable record of its inputs, execution environment, policy context, or verification path.

That gap is manageable in low-consequence experimentation and unacceptable in finance, healthcare, sovereign systems, industrial control, and autonomous operation. Compliance teams need to reconstruct not only what happened, but why a specific result was permitted to settle.

Aethelred treats provenance and policy as part of execution. The aim is to make machine-generated outputs independently inspectable without exposing all underlying data or trusting a single intermediary.

Security engineers audit validator hardware and trace an execution-evidence path through secure compute equipment
Research visualHigh-consequence outputs require reconstructable evidence of what ran, where it ran, which policy applied, and what verification passed.
Four-layer network

Execution evidence travels with the result.

Consensus

A proposed Proof of Useful Work design assigns AI jobs to validators and aggregates execution evidence through the network’s consensus path.

Confidential execution

Inference is intended to run inside attested enclaves across supported confidential-computing environments, keeping sensitive workloads isolated while producing hardware evidence.

Verification

TEE attestation is combined with zero-knowledge-backed verification. The public design names Groth16, PLONK, EZKL, Halo2, and STARK pathways rather than assuming one proof system fits every workload.

Deterministic settlement

CometBFT finality is intended to bind proof references, Digital Seals, policy metadata, transfers, block height, and validator set into a final record.

Digital Seal lifecycle

Six steps from declared intent to portable evidence.

  1. Commit

    Submit the workload definition, policy requirements, input commitments, and verification conditions before execution.

  2. Schedule

    Assign an eligible validator and supported confidential environment under the network’s scheduling rules.

  3. Attested inference

    Run the workload inside an enclave and return evidence about code identity, environment, and execution state.

  4. Proof generation

    Produce the second verification signal using a zero-knowledge or zkML-compatible proof path appropriate to the workload.

  5. On-chain settlement

    Accept the result only when both evidence paths and policy checks pass; otherwise fail closed.

  6. Digital Seal

    Issue a portable record binding the machine output to verification signals, policy metadata, and deterministic settlement.

Validator appliances connect confidential execution, proof verification, blockchain settlement, and a gold Digital Seal
Research visualThe target network connects confidential execution environments with a verifiable, policy-aware settlement layer.
Where proof matters

One trust primitive, multiple regulated domains.

FIN

Financial systems

Transaction attestation, machine-generated risk decisions, regulatory reporting, and cross-border settlement evidence.

HLT

Healthcare and research

Clinical inference trails, experiment provenance, data-handling policy, and reproducibility without exposing every sensitive input.

SOV

Sovereign infrastructure

Execution evidence, classified workflows, machine credentials, and locally governed validator infrastructure.

AUT

Autonomous systems

Decision proofs, component passports, vehicle-to-everything trust, and accountable agent-to-agent exchange.

SCM

Supply chains

Product passports, transfer integrity, customs records, and evidence for rules attached to physical goods.

AGT

AI agents

Authentication, decision-chain proofs, machine permissions, usage rights, and settlement between autonomous services.

Public roadmap

A staged path from testnet to governance-gated network.

  1. Site: complete
    Q1 2026

    Internal Testnet v1.0

    The official roadmap labels an internal network release completed.

  2. Site: in progress
    Q2 2026

    Public testnet and SDKs

    The public page describes validator onboarding, developer tooling, and SDK work in progress.

  3. Upcoming
    Q3 2026

    External security audit

    Audit, remediation, and hardening are described as a forthcoming phase, not a completed assurance.

  4. Upcoming
    Q4 2026

    Mainnet preparation

    Economic, legal, operational, and governance readiness remain prerequisites.

  5. Governance gated
    2027

    Mainnet and ecosystem

    A future objective contingent on prior evidence, review, and governance decisions.