AI hackers are coming—will cybersecurity be doomed to fail? Vitalik doesn’t agree.
His core judgment is: security naturally favors the defender. If AI can prove the Navier–Stokes equations, and Fermat’s Last Theorem, then it can also turn “this program is secure” into a mathematical theorem. The real difficulty isn’t that the code is complex; it’s how to define “security.” If the definition is too narrow, the attack surface will leak in from outside the definition.
Therefore, he argues that formal verification can’t only cover modules that people think are critical. Databases, networks, caches, and all other programs should be included; the security definition must also be readable and composable. Compared with scanning code line by line, it’s often more practical to verify first whether the definitions of components like message protocols, sandboxes, SNARKs, and FHE are sufficient.
This isn’t “good people find vulnerabilities before bad people.” It’s about making code inherently more resistant to attacks. Vitalik says continuing to put about 90% of net assets into encrypted assets is, in itself, a bet on this path. Without it, blockchains that combine scalability and privacy won’t have a future; Ethereum will move in this direction in the coming years.
#AI #网络安全 #以太坊 $ETH
His core judgment is: security naturally favors the defender. If AI can prove the Navier–Stokes equations, and Fermat’s Last Theorem, then it can also turn “this program is secure” into a mathematical theorem. The real difficulty isn’t that the code is complex; it’s how to define “security.” If the definition is too narrow, the attack surface will leak in from outside the definition.
Therefore, he argues that formal verification can’t only cover modules that people think are critical. Databases, networks, caches, and all other programs should be included; the security definition must also be readable and composable. Compared with scanning code line by line, it’s often more practical to verify first whether the definitions of components like message protocols, sandboxes, SNARKs, and FHE are sufficient.
This isn’t “good people find vulnerabilities before bad people.” It’s about making code inherently more resistant to attacks. Vitalik says continuing to put about 90% of net assets into encrypted assets is, in itself, a bet on this path. Without it, blockchains that combine scalability and privacy won’t have a future; Ethereum will move in this direction in the coming years.
#AI #网络安全 #以太坊 $ETH