For most of the modern information economy, the answer to "can we share this data?" has been one of two unsatisfying responses: share it and accept the loss of control, or protect it and accept the loss of verifiability. There is no other option, the lawyers say. There is no other option, the security architects say. There is no other option, the regulators say.
There is now another option.
I. The trust we lost
The default architecture of modern compliance is built on a stack of intermediaries — auditors, ratings agencies, regulators with subpoena power, cloud vendors with attestation chains, third-party arbiters of every kind. Each intermediary exists because verification has been expensive, slow, and required physical access to the data being verified. We have built an entire intermediary economy because we could not prove things directly.
This worked, for a while, under three preconditions:
- The volume of regulated data was small enough that intermediaries could plausibly inspect it.
- The intermediaries themselves were trustworthy.
- Data did not need to cross jurisdictional boundaries faster than treaty negotiations.
None of these preconditions hold any more. Data volumes have exceeded by orders of magnitude what any auditor can sample. Intermediaries have repeatedly proven captureable — Arthur Andersen, KPMG, Wirecard, FTX. And the modern economy moves data across borders at a velocity that no diplomatic process can govern. The result is a slow, expensive, increasingly absurd system of compliance theater, in which everyone pretends to verify what nobody can.
II. The math we found
Zero-knowledge proofs are not new. The foundational theorem — that one party can prove the truth of a statement without revealing anything about why the statement is true — was published by Goldwasser, Micali, and Rackoff in 1985. For most of the four decades since, the construction has been theoretically beautiful and practically unusable.
That has changed in the last five years. Modern PLONK-ish proof systems (Halo2, Plonky3) compose efficiently with database operators. Polynomial commitments make dataset commitments cheap. Recursive proof composition collapses long computations into single short artifacts. Hardware acceleration (GPU, FPGA, the first generation of dedicated ZK ASICs) has reduced prover cost by one to two orders of magnitude per year. The TPC-H benchmark — the canonical test for analytic SQL — now runs end-to-end in zero knowledge.
These are not incremental advances. They are the precise sequence that takes a technology from research curiosity to enterprise tooling. The same sequence that took TLS from a Netscape side-project to the substrate of every web transaction. The same sequence that took public-key infrastructure from a 1976 paper to the basis of every signed software update.
We are at the start of the sequence for verifiable computation.
The question is no longer if — it is who builds the standards, on what timeline, and for whose benefit.
III. The world we can build
When verifiability becomes a cryptographic property of the output rather than an institutional property of the producer, three things change.
Audit becomes a check, not an investigation. A central bank can verify a regulated entity's capital ratio in milliseconds, on the full ledger, with complete statistical confidence. A pharmaceutical regulator can verify a trial's primary endpoint without an inspector flying to forty sites. A national tax authority can verify a corporate return without a five-year examination. The intermediary class does not disappear — its role moves up the stack, from did this happen to should this be happening.
Cross-border data flows become tractable. Personal data stays in the jurisdiction that mandates locality. The proof — the answer — travels freely, with the certainty that no row has been disclosed. Schrems II, the Data Act, every emerging digital-sovereignty regime: the binary choice between extraterritoriality and balkanization collapses into a new, mathematical solution.
The right to verify becomes a public good. For the first time in the history of the information economy, an ordinary citizen — a journalist, a watchdog, a small business — can verify a claim made by a government agency or a corporation, without needing a Big Four firm or a regulator with subpoena power. The intermediary economy was, at its best, a centralization of the right to check. Cryptography decentralizes it.
This is the world the math now makes possible. The question is whether the institutions that build the rails will be the same institutions that built the broken ones, or whether something new is needed.
IV. What we intend to do
zkDB exists to be the institutional voice and engineering depth of this transition. We are not a SaaS vendor. We are not a research lab. We are a firm — built to engage with the regulated enterprises, central institutions, and serious public-sector bodies for whom verifiable computation is no longer a luxury.
Our work has four convictions:
One — the technology is ready, but the architecture is not. Modern proof systems are mature. The hard work now is the engineering of fit: which circuit for which query, which commitment cadence for which regulation, which key-custody design for which adversary model, which proof-system trade-off for which deployment surface. This is what we do. It is what we are best at.
Two — the language matters. Most of the existing voices in this space speak the dialect of blockchain. Enterprise CTOs and central-bank officials correctly do not want to procure a blockchain. We speak the dialect of cryptography, compliance, and engineering — the dialect the buyers actually use. The same primitives, the same math, a different vocabulary, a different audience.
Three — engagements, not products. Every regulated environment is bespoke. The CCAR submission flow of a US G-SIB and the MiFID II transaction-reporting flow of a Swiss universal bank are not the same problem. The clinical trial endpoint of a global pharma and the value-based-care contract of a US Medicare ACO are not the same problem. We build to fit, with disciplined re-use of components across engagements.
Four — institutional voice. This is a firm, not a personality. Our writing is published in the firm's name. Our briefings are conducted by our principals under the firm's name. Our research is contributed to the literature under the firm's name. The category is larger than any individual; the work must be too.
We are at the beginning. We intend to be here for decades. The next era of trustworthy data systems will be built by someone. We intend to be part of that answer.
— zkDB