Off-Chain Execution & TEE Integration

Met.AI’s architecture acknowledges the computational limitations of on-chain environments, particularly when dealing with large-scale AI inference, model orchestration, or privacy-sensitive operations. To address these challenges, the platform integrates a robust off-chain execution layer supported by Trusted Execution Environments (TEEs)—forming a hybrid computation model that ensures both performance and verifiability.

The off-chain execution framework operates through a network of registered executor nodes. These nodes are responsible for running heavy AI workloads that cannot be efficiently processed on-chain due to gas constraints, memory limitations, or runtime complexity. Each task dispatched to this layer includes encrypted input data, agent instructions, and a signed task context originating from the blockchain. Upon completion, results are returned to the chain via structured output payloads and signed cryptographic attestations.

For tasks involving sensitive or regulated data, Met.AI routes execution to TEE-enabled nodes. TEEs are hardware-isolated environments that execute code in an encrypted memory space, shielding both logic and data from host-level tampering. Within this enclave, the AI agent logic is loaded, executed, and finalized without exposing raw inputs or intermediate computations. Upon completion, the enclave generates a verifiable attestation—cryptographically signed using the chipset’s hardware root-of-trust—which is then submitted back to the chain alongside the execution result.

This architecture ensures several key properties:

• Confidentiality：User input and proprietary model logic remain encrypted throughout execution.

• Integrity：Execution follows the original, unmodified agent logic as verified by the enclave attestation.

• Verifiability：On-chain contracts can confirm that a task was processed inside a trusted environment without needing access to the underlying data.

TEE support is integrated natively at the protocol level, enabling users to flag specific tasks for secure enclave processing at submission time. Met.AI verifies node attestations using established remote attestation protocols (e.g., Intel SGX, AMD SEV) and includes challenge-response checks to prevent false reporting or proxy execution.

By combining off-chain scalability with hardware-level security guarantees, Met.AI bridges the performance gap between trustless coordination and real-world AI computation—creating a secure and extensible foundation for decentralized intelligence.