Privacy-first Decentralized AI
Maturity: Design study. This non-normative page is the public landing digest. The full architecture paper is Privacy-first Decentralized AI — Whitepaper. This is not a shipped GPU marketplace, training network, data market, or decentralized AI product.
Public site: https://gitbook.conet.network/applications/privacy-first-ai.html
The same narrative appears on conet.network and in the L0 thesis Permissionless cloud and zero-trust applications.
PRIVACY-FIRST DECENTRALIZED AI
Three independent roles. One open intelligence economy.
Centralized AI combines models, data, and user agents under one platform. CoNET separates them across independent wallet-addressed participants—and uses private application paths and programmable micropayments to make that separation economically sustainable.
Models, data, and AI agents need neither mutual trust nor ownership by the
same platform. Layer Minus and web3:// reduce the need for one network
intermediary to observe the complete user–service relationship. That
benefit is conditional on operator separation, client-side key control,
route diversity, limited identifier reuse, and the absence of collusion.
CoNET-DLE micropayments are designed so value can move among independent
participants for each measurable contribution. That is not a claim that
wallet payments or DLE automatically hide prompts, datasets, or the
service graph. The one-basis-point protocol-value fee is a target for
specified classes of applicable value movement, not a universal all-in
fee for every AI task. AI event classes are proposed, not frozen DLE
tip classes.
Design principle: Do not ask users to trust a platform that calls itself neutral. Make bias, interest, and authority visible enough to compare, refuse, and replace. A contestable model is not assumed to be unbiased. The goal is not neutral intelligence. It is an intelligence market in which no single model owner can make its bias unavoidable. Independently sourced data means data contributed under the provider’s own provenance, license, disclosure, and payment terms—not exclusively selected and preconditioned by the platform that owns the model and the user agent. Three wallets do not prove three independent powers. Role separation becomes meaningful only when ownership, infrastructure, logging, authorization, and economic interest are also separated—or when their overlap is explicitly disclosed. Decentralization is not the number of addresses. It is the inability of one control domain to reconstruct and govern the whole relationship.
Observation, prediction, and intervention
Platform analysis, prediction, and choice-shaping are related, but they are not the same capability.
| Stage | Capability | Risk |
|---|---|---|
| Observation | Collects behavioral and relational metadata | Privacy and power asymmetry |
| Prediction | Infers preference, risk, and the next action | Covert classification and differential treatment |
| Intervention | Changes ranking, price, content, or agent action | The user's choice environment is manipulated |
Prediction is not intervention. The governance risk becomes more severe when the same actor that predicts a user also controls the ranking, price, recommendation, or agent action presented to that user. Independent prediction is not automatically manipulation.
A model can be packaged and replaced as a capability. An agent is a continuing relationship: it retains context, holds authority, selects services, and mediates action over time.
Full paper: Privacy-first Decentralized AI — Whitepaper — Executive Thesis, including observation, prediction, and intervention (§1.3); the closed AI power loop; contestable data, models, and agents (independently sourced data, contestable model supply, user-controlled agents, independence as a verifiable property, and the operational evaluation matrix); three independent roles and Figure 1; private and economic coordination; protocol requirements; threat model; governance; maturity and roadmap.
Three independent roles
| Role | Line | Meaning |
|---|---|---|
| MODEL BUILDERS | Build intelligence without owning the user. | Publish models or verifiable artifacts. Do not automatically receive complete user prompts or control the data-provider network. |
| DATA PROVIDERS | Contribute data without surrendering control. | Contribute and attest inputs. Fragment and encrypt sensitive material so one collector does not become the universal data custodian. |
| AI AGENTS | Act for users without becoming the platform. | Request work, protect private context, and return results to the user. Do not silently become the model owner and data authority. |
No single role should own the model, the complete private input, and the user relationship at once. Three wallets, three companies, or three product cards do not prove that split if they share one beneficiary, one cloud, one login, one analytics store, one logging SDK, one payment address, or an exclusive contract.
PRIVATE COORDINATION
Implementation status. Layer Minus forwarding is an Implemented
capability over L0 resources. web3:// is Under development: v1
components include wallet locators, signed requests, encrypted response
correlation, Linux runtime support, and early persistent streams. Complete
cross-platform handling and general public hosting are not production
references.
Layer Minus and web3:// reduce the need for one network intermediary to
observe the complete user–service relationship.
This benefit is conditional on operator separation, client-side key control, route diversity, limited identifier reuse, and the absence of collusion. Protocol-role separation alone does not prove legal-entity or infrastructure independence.
- In the proposed application model, the durable application owner is
represented by a wallet identity. Layer Minus routing additionally
depends on OpenPGP identity and mailbox-route bindings; an exact
@BeamioTagmay provide an application-facing alias. Wallet addressing is pseudonymous, not automatically anonymous. - Entry, mailbox, and application hosts must be different control domains. One company can still run A, B, and C.
- Users must not reuse one public wallet for every model, data, and payment task. A stable wallet can recreate the relationship graph.
- Entry, mailbox, and hop nodes carry ciphertext. They are not the product platform.
- L0 GPU or other compute hosts remain untrusted. Encrypted fragmentation, independently controlled keys, and verifiable or redundant execution stay application-layer work.
- Applications should encrypt and fragment sensitive state so that no single storage provider receives enough material to reconstruct the whole. Fragmentation thresholds, redundancy, recovery, deletion, and availability verification remain application-specific requirements. Layer Minus provides forwarding and mailbox primitives; fragmentation is not an automatic property of all L0 traffic.
See How to use Layer Minus and the
web3:// Application Protocol.
ECONOMIC COORDINATION
CoNET-DLE is designed for small, frequent payments among agents, models, data providers, and infrastructure operators. Each participant can earn from a measurable contribution without controlling the entire AI stack.
The current DLE economic design targets a one-basis-point protocol value fee for specified classes of applicable value movement. This is not a universal all-in fee for every AI task. Inference execution, data access, proof verification, L1 settlement, retries, storage, and availability may carry separate measured costs under future class-specific fee schedules.
AI event classes are proposed, not yet normative
Examples such as one model inference, one licensed data access, or one verification result are candidate AI-market events. They are not presently frozen DLE tip classes. Each would require its own deterministic state machine, evidence format, replay rules, completion criteria, dispute process, privacy boundary, and release gates.
Official developer documentation still lists DLE as a normative design and experimental environment, not a production SDK or a launched L2 service. See L2 development.
Privacy-preserving settlement requirement
A future AI payment protocol should prove only the service right, amount, and settlement condition required for one interaction. It should avoid placing prompts, dataset identifiers, long-term agent identity, or the complete service graph on a public ledger.
This property requires a specified payment-session and unlinkability design; it is not provided automatically by wallet payments or by DLE.
A public-chain micropayment can add metadata rather than hide it.
See L2 — CoNET-DLE and L2 economics.
Closing
Privacy creates independence.
Micropayments make independence sustainable.
What this page does not claim
- A general GPU marketplace, training network, or data market is in production.
- That Layer Minus provides anonymity, or that A/B/C role labels prove legal-entity or infrastructure independence.
- That protocol-role separation alone hides the complete user–service relationship without operator separation, limited identifier reuse, and the absence of collusion.
- CoNET-DLE already settles AI micropayments in production, or that listed AI events are frozen DLE tip classes.
- That the one-basis-point protocol-value fee is a universal all-in fee for every AI task.
- That wallet payments or DLE automatically hide prompts, dataset identifiers, long-term agent identity, or the complete service graph. Privacy-preserving settlement is a specified design requirement, not a live capability.
- CoNET hosts a globally unbiased dataset or a zero-bias model.
- Decentralized AI is a live CoNET product name.
- Any listed Guardian node is an honest compute or data host.
- That
web3://is a complete production cross-platform handler or general public hosting service. - That Layer Minus application identity is only a wallet address.
- That all L0 traffic is automatically fragmented so no node holds a reconstructable whole.
- That every predictive analysis is manipulation. Prediction and intervention are distinct stages.