Page: Integra's Three-Tier Architecture URL: https://whitepaper.integralayer.com/integra-the-real-estate-layer-1/integras-three-tier-architecture Change type: Edit — ground AI Security Layer claims, clarify what's design vs deployed
Accurate description of Cosmos/EVM, DPoS, Ethermint. Keep as-is.
Replace with:
Agent Scope Enforcement
Integra's agent authorization model defines what an agent is permitted to do — asset classes, geographic markets, maximum transaction sizes, permitted counterparty types. When an agent submits a transaction, the protocol validates it against the agent's authorized scope before execution. Transactions outside scope are rejected at the contract level, regardless of what triggered them.
This provides a hard boundary against agent misbehavior: even if an agent's off-chain logic is compromised, manipulated, or produces unexpected outputs, the on-chain authorization layer enforces the principal's constraints. The agent cannot spend funds, trade assets, or interact with contracts beyond what its principal explicitly authorized.
AI Output Validation
For regulated parameters (compliance status, jurisdiction checks, transaction limits), deterministic on-chain checks override AI-generated outputs. Agent reasoning and language generation execute off-chain; only structured, schema-validated outputs are submitted for on-chain execution.
Current text is accurate. Add one clarifying sentence:
"Agent operators are responsible for their own inference infrastructure. Integra does not host or operate agent execution logic — the protocol provides identity, authorization, and settlement infrastructure that agents plug into."
Fee grants, commission routing, and stablecoin settlement are accurately described.
List of applications is accurate.