AI in Web3: Use case or buzzword? That was the big question at IBW, where CertiK CBO Jason Jiang joined a panel to explore the real opportunities and risks of AI agents in DeFi. AI is already here—from trading bots to audit assistants—and with it comes a new kind of unpredictability. Unlike smart contracts, AI agents can act differently given the same input, adding complexity to how we assess risk. Jason shared how CertiK has begun integrating AI tools internally, and how our researchers are already uncovering interesting behaviors in the field. He also outlined the dangers of trusting autonomous agents too blindly: • Prompt injection and input spoofing • Bugs in AI-generated smart contracts • AI agents reacting to fake on-chain signals The takeaway: AI logic is still code. And just like any other code that moves assets or deploys contracts—it needs to be audited.