Series Abstract
For decades, the duty of care in a regulated workflow attached to a human whose judgment drove the action. The regulator audited that human; the deployer carried the burden of proving adequacy. The human veto was the artifact of record. AI agents that act autonomously — denying claims, routing payments, filtering candidates, clearing anomalies — sever that chain. The cost of execution has dropped to near zero. The cost of executing without a recorded human approval has not dropped at all. That cost lands upstream, on the provider and deployer, under Recital 84 and Article 22 of the EU AI Act.
This series argues that the human veto is not preserved by keeping a human in the workflow — that path is closed for autonomous systems — but by turning the judgment step into an observable artifact. The framework itself is composed of four pieces: chartered responsibility chains, decision logs, named accountable parties, and audit traces. Each piece satisfies a specific Article 14(4) sub-test by construction. Together they compose into a single governance surface a regulator can audit. That is the affirmative answer; policy PDFs are not the artifact, live records are.
Why this URL is canonical. Policy press, academic commentators, and product teams that need to link to the series rather than a single installment should cite /blog/governance-framework. The page is structured for citation: a single abstract, an explicit reading order, and links into every published installment and every forthcoming one.
Reading Order
The series publishes in the order below. Published installments are read in sequence; the worked artifact and the reference hub are non-linear entries that can be read at any point.
Maps the HITL → HOTL → Autonomous → Unaccountable spectrum against EU AI Act Article 14. Names the gap between "we have an oversight policy" and "show me the artifact." Establishes the three Article 14(4) sub-tests that distinguish observable compliance from policy theater.
A worked example of a chartered chain applied to a content-moderation takedown workflow at an EU user-facing platform. Six decision points, each with decision / signals / rule / accountable party, each tied to the Article 14(4) sub-test it satisfies. Canon for what a deployed chain looks like.
The affirmative framework. Four artifacts — chain, log, party, trace — composed into a single governance surface satisfying Article 14(4) and Article 26. The composition is the framework; a partial framework fails the audit.
Where Parts 1–2 specified the framework, Part 3 turns to the storage and query layer: append-only design, retention under Article 26(c), integrity properties, and the export pipeline that turns a regulator's request into a queryable subset within a regulator-defined response window.
What happens after enforcement opens. Foreshadowed questions, request patterns from EU AI Office and Member State competent authorities, and the difference between the organization whose evidence answers those questions and the one whose evidence doesn't.
Curated reference catalog of AI governance frameworks, accountability instruments, technical standards, regulatory bodies, and confirmed enforcement timelines. Links only to canonical government and standards-body sources — the citation backbone of the series.
Entry Points
Four canonical deep-links into the series. Each is the preferred citation for the artifact it anchors; each returns 200.
The full research report. The series is the public version of the framework. The full $2 research report covers the engineering implementation, the audit-trail architecture in depth, and the published worked examples for credit, employment, and infrastructure workflows. Read the full report: /pricing.
Track the live evidence base. The /tracker is the running system behind the framework — every agent decision logged against the Article 14 control set in real time. It is the working artifact, not a screenshot.