[Editorial] The Future of DeFi Depends on the ‘Invisible Brain’

DeFi (decentralized finance) is now the core infrastructure supporting the cryptocurrency ecosystem. The total value locked (TVL) in DeFi platforms currently exceeds $100 billion as of May 2025, which means that more than half of all blockchain activities are directly related to DeFi. Almost all traditional financial functions such as exchanges, lending, asset management, and derivatives are now automatically executed on the blockchain. Now, artificial intelligence, not humans, is responsible for judging and executing these complex financial activities. Wallets are making investment decisions on their own, and AI is managing funds in DAOs (decentralized autonomous organizations). This is not just a simple technological convergence, but a paradigm shift that changes the subject and responsibility structure of the financial system. However, this transformation conflicts with the basic design of blockchain. Blockchain is a structure where all data is public. Transparency was synonymous with trust. But when AI agents become the main actors, this transparency becomes a weakness. If the AI's judgment criteria, strategic logic, and execution conditions are recorded on-chain, there are risks of imitation and front-running, and the possibility of system-wide "hive mind" increases. A predictable artificial intelligence can no longer be called "intelligent". To solve this problem, the concept of "Private Brain" has recently gained attention. This method involves hiding the decision-making process and only recording the results on the blockchain. The key technology enabling this is TEE (Trusted Execution Environment). TEE performs computations in an isolated area within the CPU and completely blocks external access. AI develops strategies in this space, and the blockchain only receives and verifies the results. Internationally, this trend has already entered the practical stage. Oasis Labs has built a TEE-based AI computation platform, Secret Network executes smart contracts while hiding input values and logic, and Cartesi and ARPA are realizing the structure of "invisible judgment, verifiable results" through computational concealment technology. However, domestically, there is no institutional preparation for these changes. Neither the Digital Asset Act nor the Electronic Financial Transaction Act defines the legal status or responsibility structure of AI agents. Key technologies like TEE, MPC, and ZK (zero-knowledge proof) remain outside the institutional framework. Innovation is accelerating, but laws and institutions are trapped in the past. Now is the time to change direction. First, the legal status and responsibility structure of AI agents must be clarified. Second, computational concealment technologies like TEE must be provided with special provisions and technical standards to operate safely within the law. Third, a "trusted computation certification system" should be introduced to verify the reliability of technology. Fourth, a digital sandbox should be promptly prepared to allow AI agents to be tested in actual industries such as the National Pension, public data, and energy trading. DeFi is not just a trend. It is currently the main system practically supporting the cryptocurrency ecosystem, and within it, AI agents are leading the "next evolution". Technology has already made "invisible judgment" possible. Now we need to be prepared to accept that judgment. Trust no longer depends only on what is visible. If it can be proven even when not visible, that is stronger trust. And only on that trust can the Republic of Korea establish its digital sovereignty.