Project summary

I am working on this important issue: if an advanced AI system behaves dangerously after deployment, who can actually stop it?

It is easy to say that a system will have human oversight or a shutdown mechanism. It is harder to identify the person or institution with the legal authority, technical access, contractual rights, information, and operational capacity to intervene. Those powers may be divided among the developer, deployer, cloud provider, government agency, regulator, and vendor. In a serious incident, nominal human oversight may prove meaningless.

I want to spend three months turning my existing research into a framework that policymakers, developers, vendors, and high-stakes deployers can use to test whether intervention authority is real before a crisis occurs.

What are this project's goals? How will you achieve them?

I will examine how legal, contractual, institutional, and technical authority is distributed after an AI system is deployed. I will conduct targeted interviews with policymakers, researchers, vendors, deployers, and civil-society experts, including participants in the United States and Africa.

The principal outputs will be:

1. A research paper explaining where intervention authority can break down.

2. An authority map showing who can detect, direct, restrict, suspend, or terminate an AI system.

3. A six-part test for determining whether claimed human control is meaningful.

4. Scenario-based testing of the framework.

5. Model policy and procurement language for institutions acquiring or deploying consequential AI systems.

6. A public checklist that other researchers and practitioners can criticize, test, and improve.

The project could reduce catastrophic risk by helping institutions identify and correct gaps in control before advanced systems are widely deployed. A technically corrigible system may still be practically uncontrollable if nobody has the complete authority and capacity to intervene.

How will this funding be used?

The first $15,000 would fund approximately two months of protected research time and allow me to complete the core research paper, authority map, initial interviews, and evaluation framework.

Funding up to $24,000 would support three months of full-time work, broader stakeholder interviews, additional scenario testing, model policy and procurement language, and public dissemination. The $24,000 includes the taxes, healthcare, and living costs necessary for me to pause unrelated document-review work.

Funding between $24,000 and the $35,000 maximum would support independent technical and legal review, interview transcription or research assistance, and travel for international stakeholder engagement and presentation of the findings, including engagement with African AI-governance communities.

I have also applied to BlueDot Impact and the Transformative AI Fund for overlapping support. I will disclose any award and will not accept duplicative funding for the same time or expenses.

Who is on your team? What's your track record on similar projects?

I will lead the project as an independent researcher. I do not currently have paid staff.

I am an attorney, educator, and former congressional Chief of Staff with more than twenty years of experience across Congress, the executive branch, technology-policy advocacy, international trade law, consulting, and university teaching. I spent eleven years as Chief of Staff and Chief Counsel to a Member of Congress, where I managed 26 employees and a $1.6 million annual office budget. I also served as Vice President of Public Policy at the Computer & Communications Industry Association and as a federal attorney at the U.S. Department of Commerce.

My recent AI-governance work has appeared in FedScoop, Nextgov, Tech Policy Press, RealClearDefense, Techpoint Africa, and Broadband Breakfast. I have presented versions of this research through the AI Sovereignty Network and at N-SEA 2026, and related work has been accepted for HCOMP 2026 and other academic and policy conferences.

I am also Director of Policy and Research at AI Safety Nigeria and an Inner Council Member of the AI Sovereignty Network. In less than three months, I grew my independent AI-governance Substack to approximately 400 subscribers.

What are the most likely causes and outcomes if this project fails?

The largest risk is that the framework remains too conceptual and does not receive enough technical criticism or real-world testing. I may also have difficulty recruiting a sufficiently diverse set of interviewees within three months. Even if the work is completed, policymakers, vendors, and deployers may not immediately adopt it.

I will reduce these risks by publishing an early version, seeking criticism from both technical and policy experts, testing the framework against specific deployment scenarios, and producing a short practical checklist in addition to the longer paper.

A partial failure would still produce a clearer account of the gaps between nominal oversight and actual intervention authority. The worst likely outcome is a useful paper and preliminary framework that receives limited adoption and requires further testing.

How much money have you raised in the last 12 months, and from where?

I have raised $0 for this project during the last 12 months and have been developing it through unpaid, self-funded work.

I applied to the BlueDot Impact Rapid Grant on September 6, 2026, requesting $24,000; the application is pending. I also applied to the Transformative AI Fund on September 6, 2026, requesting $24,000; that application is pending. My Round 1 application to the Corrigibility Research Fund was declined on September 4, 2026, and I received no funding.

These applications overlap with the core three-month project. I will disclose any award, decline duplicative funding, and adjust the scope of other requests as necessary.