When South Korean President Lee Jae-myung steps into the San Francisco AI Summit this week, flanked by meetings with the CEOs of Nvidia, OpenAI, Anthropic, and Broadcom, the world sees a nation securing its place in the AI arms race. But from my vantage point—a cryptographer who has spent years auditing the soul of code—this is not a victory lap for centralized technology. It is a red flag. The fact that a head of state must personally negotiate for access to black-box models and proprietary chips tells us something profound: trust is not being built; it is being concentrated.
Behind the diplomatic gloss lies a stark reality. Korea, a global leader in semiconductor manufacturing, is now reduced to a supplicant for the very tools its factories helped enable. The meeting list is telling: Nvidia for hardware, OpenAI and Anthropic for frontier models, Broadcom for the networking fabric. Missing is any mention of decentralized infrastructure, of open-source alternatives, of blockchain-based verification. This omission speaks volumes about the current trajectory—a trajectory where power aggregates in a handful of Silicon Valley boardrooms. For those of us who lived through the chaos of 2017 and the idealism of DeFi Summer, this feels like a regression. We forged a compass then to navigate decentralization, but the compass needle now points toward a new kind of centralization: algorithmic feudalism.
Yet within this pilgrimage lies an unspoken opportunity. The cryptographic community has long theorized that AI and blockchain are natural allies. After the Dencun upgrade, we saw blob data usage spike, and I have argued that within two years, post-Dencun blob space will be saturated, driving rollup gas fees up again. But more fundamentally, the core challenge of AI—verifiability of model behavior, provenance of data, and accountability of decisions—maps precisely onto blockchain's strengths. In my work on the Human-Centric AI Ledger, I developed a cryptographic protocol that allows any AI output to be linked back to the model version, input data, and inference parameters via a tamper-proof on-chain hash. This is not science fiction; it is a practical extension of the zero-knowledge proofs we use in zk-rollups today.
Imagine a future where President Lee does not just meet with closed-source providers, but also with the teams building decentralized AI marketplaces—projects that use token incentives to reward high-quality training data, that employ on-chain governance to audit model updates, that allow communities to stake tokens against the integrity of AI outputs. The meeting list should have included a blockchain-based AI verifier. The fact that it did not reveals a blind spot in current statecraft. The same moral-first cryptographic audit that I applied to ICOs in 2017 now applies to AI models. Trust is not a metric you can buy; it is a memory we share. And memory, in the digital age, must be recorded on an immutable ledger.
A skeptic might argue that blockchain adds unnecessary overhead to AI systems, that latency and cost make it impractical for real-time inference. That the Korean government is right to prioritize raw performance over philosophical purity. But this view ignores the long-term cost of opacity. When an AI model denies a loan, recommends a medical treatment, or controls a power grid, the lack of auditable reasoning creates systemic risk. We saw in 2022 how misaligned incentives in DeFi led to catastrophic failures. The same will happen in AI if we do not embed accountability at the protocol level. The contrarian truth is that blockchain is not an enemy of AI efficiency; it is the only mechanism we have to prevent the AI giants from becoming the new too-big-to-fail institutions. From the chaos of 2017, we forged a compass—and that compass points toward decentralized verification, not centralized convenience.
President Lee's summit is a reminder that the convergence of AI and crypto is not a niche interest; it is the defining infrastructure challenge of our decade. The question is whether we, as a community, will build the ethical guardrails before the next crash. Or will we wait until trust is not a memory, but a broken promise? The choice is ours.
