MetaMask Agent Wallet authorization scope creates DeFi financial risk
MetaMask launched an Agent Wallet in June 2026 that lets AI agents execute DeFi transactions under user-defined authorization rules. Experts warn that broad authorization scopes, user oversight failures, and the programmable nature of wallet permissions create a new risk layer where agents could drain funds beyond intended limits, despite protective safeguards like transaction simulation and spending caps.
What happened
MetaMask launched early access to Agent Wallet on June 8, 2026, enabling AI agents to execute DeFi transactions including swaps, perpetuals, liquidity provision, and other on-chain workflows across EVM chains and Hyperliquid.
The wallet requires users to define authorization rules in advance through spend limits, allowlists, transaction simulation, threat scanning, MEV protection, and two-factor approval for flagged transactions. Base separately rolled out a Model Context Protocol gateway enabling AI models to connect to user accounts and execute DeFi actions via OAuth 2.1.
Reported impact
The primary risk identified is authorization scope: agents could drain more funds than intended if users grant broad permissions and fail to revoke them, even without a protocol bug or exploit. One scenario describes an agent authorized to rebalance across Uniswap and supply collateral on Morpho, then draining more than expected weeks later due to forgotten revocation.
Index analysis
The shift from human-signed transactions to agent-executed workflows creates a new risk layer where wallet authorization and permission design become the primary attack surface. Account abstraction and smart account technology (30 million ERC-4337 accounts live by June 2026) enable programmable permissions and session keys, but also programmable authorization boundaries that could be exploited.
MetaMask's protective safeguards—transaction simulation, threat scanning, MEV protection, and $10,000/month Transaction Protection coverage during Early Access—attempt to detect anomalies before signing and provide a limited backstop if protections fail, but experts question whether these rules make agentic DeFi materially safer or create a more programmable attack surface.
Governance controls
MetaMask Agent Wallet implements mandatory simulation, threat scans, MEV protection, and limited coverage insurance before agent-sent transactions. Base's OAuth 2.1 framework governs access with action-specific plugin affordances to limit scope. Users retain control of private keys while agents operate under agent-specific wallets.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.