BlockBeats News, April 16th, Aave Labs announced the launch of the AI-driven governance security system Aave Checkpoint, adding a structured multi-layer review process to each proposal and payload before execution. The workflow of the Checkpoint is divided into two stages:
Automatic AI Analysis Stage: When a new proposal is submitted, the system automatically fetches on-chain data, proposal source code, and IPFS text, cross-references parameters with governance specifications, uses Seatbelt to simulate output and check the proposal’s execution path, runs over 60 specialized security analysis agents for vulnerability detection, DeFi-specific risk analysis, agent upgrade security checks, etc., and generates a structured audit report with pass/fail conclusions;
Mandatory Human Review Stage: Each report generated by AI must be independently signed by at least two individuals before it is considered review completed. Checkpoint covers the Aave V3, V4, GHO, and Aptos-v3 protocols, including 5 custom Aave governance skills and 52 external security skills.
