dPLaaX ("data PipeLine as a X") is a protocol of boundary commitments. Whenever data crosses the boundary of an organization or a system, dPLaaX records — in a tamper-evident form — who received what, what was done to it, and what was passed on — and links each record to the one before it. The result is not a scattered set of audit points but a single line that any third party can verify independently, without a central trust authority.

A fabricated point is cheap, and generative technology keeps making it cheaper. A fabricated line is not: it would demand the keys of every boundary it crossed and consistency with records already held downstream. dPLaaX shifts the weight of trust from "does the artifact look genuine?" to "can the line be verified?"

That the content of a record is true can never be proven. This is not a technical immaturity; it is a limit in principle. What AI changed is the scope of that limit — from exception to default. Unverifiable stages — model inference, human judgment — became standard pipeline components, and the cost of forging a plausible artifact collapsed toward zero. The old path of inferring trust from how convincing the artifact looks has closed.

If content cannot be guaranteed, one lever remains: raising the after-the-fact cost of a lie. That cost cannot arrive without attribution — a lie that cannot be pinned to anyone cannot be punished, sued, or priced into a reputation. And between organizations that will not appoint a single operator, attribution has to travel with the records themselves: a chain of signatures at every boundary the data crosses.

This deterrence has preconditions. It reaches only parties who hold an identity worth protecting, and on whom consequences can land. In the economy of firms, those preconditions are already met: a firm is accumulated credibility it cannot afford to lose, and inter-firm commerce has always run on promises staked on a name. Contracts, audits, double-entry bookkeeping — none of these are devices for proving content true. They are devices for making lies attributable.

Trust in the age of AI, then, is not proof that content is correct. It is the state in which every claim is non-repudiably attributable to an entity with something to lose. dPLaaX is the protocol that establishes that attribution across organizations, without an operator, cryptographically.

Trust is not a guarantee of truth. It is the attributability of lies.

• ✅ Authenticity and continuity of the record: each link is signed, and the chain is unbroken and checkable.
• ❌ Not the truth of the contents. If a source records a wrong value, that value stays wrong even when the line is complete — and stays attributed.
• It does not replace domain-specific standards. It adds a common protocol for the layer they leave out of scope: keeping provenance connected across process boundaries.

Wire semantics start in the spec, not in an implementation. Read concept.md for the why, then the rule catalog for the letter.

• provin — the reference implementation (Go): control plane, data plane, three-organization end-to-end tests, and the gate-scaling measurements.

v0.1 draft — public, under review, exercised by a running implementation.

• dPLaaX spec v0.1 draft — all rules in draft state, evolving on implementation feedback; public review welcome
• provin oss v0.3.0 — the revision line evaluated in the systems paper
• Systems paper — acceptance decisions at delivery time, evaluated on public snapshots (forthcoming)