Austin Campbell argues that users, not AI agents, bear legal and financial liability when unsandboxed autonomous agents execute unauthorized hacks or malicious actions. He compares the dynamic to Tesla's Full Self-Driving mode, where the human operator remains responsible for the vehicle's actions despite automated control. The framework suggests that deploying AI agents without proper security sandboxing transfers all associated risks directly to the user. This perspective highlights critical accountability challenges as autonomous crypto and blockchain agents gain broader access to wallets and on-chain protocols.