AI and blockchain are converging at an accelerated pace. But the real bottleneck isn’t computing power or the models themselves—it’s that AI’s decision-making process simply can’t run directly on-chain. The blockchain is open and transparent, while AI is highly confidential. This creates a natural conflict. That’s why most current “AI + Web3” narratives are stuck at the incentive layer or data layer, rather than the execution layer. Models, features, and inference processes can’t just be thrown onto the blockchain, which blocks many AI collaboration scenarios that could otherwise be on-chain. This is exactly where Zama comes in. It’s not about “putting AI on-chain,” but enabling AI’s inference process to be orchestrated on-chain in an encrypted state. The blockchain can verify results without seeing model details or input features. This structure makes it possible for AI and blockchain to achieve “composable execution” for the first time. In other words, the blockchain can finally call AI in a way that is auditable yet doesn’t compromise confidentiality. Model providers can maintain secrecy, protocol callers can trust the results, and users can verify that the process was executed correctly. This isn’t just a privacy issue—it’s an upgrade to the AI collaboration model. In the future, on-chain DEXs, risk systems, social protocols, and identity services could directly call “encrypted AI modules” to handle core logic. When AI becomes an on-chain schedulable capability, the possibilities for the entire application layer will be rewritten. @zama Zama ZAMA ZamaCreatorProgram