Global Accountability Initiative · Introductory briefing

Verify power.
Preserve freedom.

Prepared by Jeff Owen, private citizen · November 8, 2026
Working introduction accompanying Revision 3 · For discussion and critique

Why I am asking for your attention

Each of us has a limited life. During that time, we depend on shared rules to protect our freedom and give us a fair chance. When people or institutions gain an unlawful advantage by bending those rules out of sight, the consequences fall on others who may have little ability to discover or challenge what happened.

I want to explore whether AI can make a practical contribution to addressing that imbalance now. The purpose is to help people examine evidence and strengthen fair accountability, while preserving legitimate privacy and preventing the creation of another unaccountable power.

The Global Accountability Initiative is a citizen-led discussion proposal. It is not an operating watchdog, an allegation against any person or institution, or a claim that AI can determine guilt. Its value must be established through scrutiny and evidence.

A specific question worth examining

The proposed starting point is a bounded public-procurement study: can AI-assisted review improve the quality or efficiency of examining public purchasing records compared with conventional methods and expert review, without unacceptable harm?

Potential tasks include extracting information from documents, connecting relevant records, and preparing evidence for human assessment. These are hypotheses to test. A persuasive AI explanation, an unusual pattern, or agreement between models would not establish misconduct.

The study would focus on one jurisdiction or clearly defined group of agencies, a manageable procurement question, and lawfully available public records. It would compare results and total effort with credible alternatives, including the burden of checking false leads and correcting mistakes.

Build on what already exists

Before proposing a new institution or claiming novelty, the initiative calls for an audit of existing tools, research, standards, and organizations. The useful outcome may be contributing to an established effort, independently evaluating existing methods, building a narrow missing component, or deciding that this proposal adds too little to proceed.

1 / 2 · Discussion proposal · Participation in review does not imply endorsement.

Purpose, boundaries & invitation

Accountability must apply to the tool itself

The proposal makes privacy, evidence quality, human responsibility, and independent correction central to the research design. The immediate pilot would not access private banking records, impose penalties, or automatically refer people for enforcement. Research would remain nonpublic by default; any proposed public naming would require a separate, justified and independently reviewed decision.

Evaluation would examine investigator workload, reliability of evidence, errors in matching identities, unequal impacts, costs, and the ability to correct harm. Negative findings should be published. Failure to demonstrate sufficient value, or inability to maintain necessary safeguards, should lead to narrowing, redesign, or stopping.

The document also considers a possible long-term federated network. That is optional and separate from the pilot. Supporting the procurement research question does not require accepting the network, a new organization, or future expansion into protected data.

What remains unresolved

The draft does not claim that a suitable jurisdiction, independent funding, workable governance, or an advantage over existing methods has been established. It includes a targeted source check with citations, distinguishes supported premises from unresolved claims, and identifies questions that need procurement, legal, statistical, civil-liberties, and technical expertise.

AI-assisted drafting has helped organize the proposal and expose objections. Agreement among models does not validate the idea. Independent human examination is the next meaningful step.

What I am requesting from your organization

I would welcome a brief assessment, relevant references, or an introduction to someone qualified to address these questions:

  1. Is this a useful and answerable research question? What would make it worth testing?
  2. Which existing projects, datasets, or evaluations should it join, build on, or learn from?
  3. Which legal, practical, methodological, or rights-related weaknesses could make it unworkable?
  4. What should a credible first study measure, and who should independently evaluate it?

The request at this stage is for critique and feasibility guidance. Your participation would not imply endorsement. I would seek permission before representing an organization or reviewer as a supporter.

Thank you for considering whether this idea merits further examination.

Jeff Owen
Private citizen · [email protected]
globalaccountabilityinitiative.org — full discussion document