AI Agents with Wallets: What Could Go Wrong?

​When an AI agent is tasked with managing digital assets, over-optimization without human wisdom can lead to costly on-chain blunders in seconds. Even worse, trying to give private key to an agent always end up with a leak. Through real-world experiments and surprising failure modes, this talk breaks down why traditional wallet security falls short for non-human actors. Philippe Hébrard, Head of Product at Ledger, unpacks how to bridge the gap between machine autonomy and cryptographic trust before handing AI the keys to the kingdom.