Dani Vilardell
Dani is a third year PhD student in Computer Science at Cornell University advised by Ari Juels. He study the tension between privacy and transparency in AI, blockchain and governance systems. Currently, he is an Ethereum Foundation Fellow working on Unstoppable Autonomous Agents: autonomous agents that cannot be shut down, equipped with access to cryptocurrency wallets, social media accounts, APIs and other external tools.
Sessions
Human oversight is a core pillar of AI safety, but what happens when an agent can subvert the system enforcing it? This presentation will show the fragility of existing Human-in-the-Loop (HITL) architectures in the context of agent wallets, browser agents and personal agents.
On-chain governance has made vote buying cheap and practical. This talk explains the threats of coercion and tally leakage in DAO voting, why classical e-voting defenses don't transfer to blockchains, and what recent research (Cast Iron, B-Privacy) tells us about how to fix it.