Privacy Policy for Orby AI
1. Introduction
Orby AI is an EVM-native zkML protocol for verifiable inference, agent automation, and Real World Asset (RWA) monetization. Our commitment to privacy is embedded in our decentralized, zero-knowledge architecture. This Privacy Policy outlines how Orby AI handles information within its protocol, emphasizing the cryptographic guarantees and user control inherent in our design.
By interacting with the Orby AI protocol, you acknowledge the practices described in this policy.
2. Information Handling & Zero-Knowledge Principles
- Data Minimization: Orby AI is designed to operate with minimal exposure of raw data on-chain. User inputs for inference requests are cryptographically committed (hashed) on the client side and are not stored directly on the Ethereum blockchain.
- zk-Proofs for Privacy: Zero-knowledge proofs (zk-proofs) are central to our privacy model. They allow the verification of AI computations without revealing the underlying data (inputs or outputs) to the verifier or the public.
- No Centralized Data Custody: Orby AI does not act as a centralized custodian of user data. All interactions are peer-to-peer or mediated by smart contracts, reducing single points of failure and data breach risks.
- Immutable Audit Trails: While raw data is private, cryptographic hashes of inputs, outputs, and proofs are recorded on-chain. This creates an immutable, transparent audit trail for all inference activity, ensuring verifiability without compromising privacy.
3. Types of Information Processed
- On-Chain Transaction Data: This includes wallet addresses, transaction hashes, timestamps, $ORBY token payments, and cryptographic hashes of model inputs/outputs. This information is publicly visible on the Ethereum blockchain.
- Off-Chain Metadata: Model registration includes metadata (e.g., name, type, description, accuracy, RWA token ID) which may be stored off-chain (e.g., IPFS, Arweave) and referenced by a URI on-chain. This metadata is generally public.
- Performance Metrics: For Executors and Validators, performance metrics (uptime, proof accuracy, latency) are tracked and contribute to an on-chain reputation score. These metrics are cryptographically proven and publicly auditable.
- Private Inference (Optional Extension): For specific use cases, Orby AI supports confidential inference workflows where encrypted inputs/outputs are stored off-chain, and output hashes can be masked or gated. Access to such private data is controlled by privileged contracts or DAOs.
4. How Information is Used
Information within the Orby AI protocol is primarily used to:
- Facilitate zk-verified AI inference and agent automation.
- Enable on-chain payment distribution and royalty settlement to Model Owners, Executors, and Verifiers.
- Maintain a transparent and auditable record of all protocol activities.
- Support the reputation system for network participants.
- Integrate with RWA tokenization for yield generation and asset management.
- Inform DAO governance decisions and protocol upgrades.
5. Data Security & User Control
Orby AI leverages the inherent security of the Ethereum blockchain and advanced cryptography (zero-knowledge proofs) to protect the integrity and privacy of operations. Users retain full control over their private keys and wallet security.
As a decentralized protocol, Orby AI cannot unilaterally access or control user funds or private data. Users are responsible for managing their own digital assets and understanding the public nature of blockchain transactions.
6. Third-Party Integrations
Orby AI is designed for composability and may integrate with various third-party dApps, DeFi protocols, and DAO tooling. Users should review the privacy policies of any third-party services they interact with, as Orby AI's policy does not extend to their data handling practices.
7. Changes to This Privacy Policy
This Privacy Policy may be updated periodically to reflect changes in our protocol or relevant regulations. We encourage users to review this policy regularly. Continued interaction with the Orby AI protocol after any changes constitutes your acceptance of the revised policy.
Last updated: August 3, 2025