Skip to content
Sprint projectJun 21, 2026Lagos, Nigeria

Closing the Sovereign Safety Gap: A Localized Adversarial Auditing Framework for African AI Governance

Omonivie Cynthia Jatto · Team LAAI

Submitted to Global South AI Safety Hackathon. Sprint projects are early-stage work by participants, not Apart Research publications.

Read the report

Report: Closing the Sovereign Safety Gap: A Localized Adversarial Auditing Framework for African AI Governance

Share

The Localized Adversarial Auditing Initiative (LAAI) proposes building African-led technical audit capacity for AI systems deployed on the continent, drawing on offensive security methodology rather than purely academic evaluation design. Nigeria ranks 35th globally on Policy Capacity but only 72nd overall on the 2025 Oxford Government AI Readiness Index, a gap that holds across Sub-Saharan Africa's top-ranked countries and shows that policy design is outpacing technical verification capacity. This report presents the LAAI audit framework, a pilot case scoping for an African healthcare AI deployment, and the policy case for why technical evaluation capacity, not compute, is Africa's most realistic governance leverage point.

Reviews

Judging this Sprint?

Review this project

Your public critique appears on this page without your name. Your private critique is not published; only the Apart team reads it. If you agree below, we share your review with grantmaking.ai (opens in new tab) and the Transformative AI Fund so strong projects can be funded.

Not shown on this page.

Shown on this page, without your name.

Only the Apart team reads this, and funders if you agree below.

Share my name publicly on grantmaking.ai *
Share my private critique with funders *

How much would this matter for AI safety if it worked? How innovative is it? For scores of 4-5: is this actually new to the field, or replicating recent work?

Scoring guide
  1. 1Negligible. No clear problem addressed, or no meaningful novelty.
  2. 2Limited. Addresses a real problem but with a generic or well-trodden approach. Incremental at best.
  3. 3Moderate. Clear problem with a reasonable approach; some novelty in framing or method beyond routine application of existing tools.
  4. 4Significant. Important problem with an original approach, or identifies a neglected problem area. A valuable contribution others could build on.
  5. 5Exceptional. Tackles a critical AI safety problem with a genuinely novel approach, or opens a new research direction. Clear theory of change. You'd be excited to share this with researchers in the area.

How sound are methodology, implementation, and findings?

Scoring guide
  1. 1Seriously flawed. Methodology broken, results uninterpretable, or implementation doesn't work.
  2. 2Weak. Approach has significant gaps: missing validation, flawed experimental design, or incomplete implementation.
  3. 3Competent. Technically solid given the short duration. Methodology makes sense, results are interpretable, limitations acknowledged, work builds toward clear conclusions.
  4. 4Strong. Thorough methodology with convincing validation. Results clearly support conclusions. Immediately useful for future work.
  5. 5Exceptional. Ambitious scope executed rigorously. Surprising findings, novel methods, or unusually robust validation.

How clearly are work, findings, and impact potential communicated?

Scoring guide
  1. 1Incomprehensible. Cannot determine what the project is actually claiming or doing.
  2. 2Hard to follow. Key information buried, missing, or diluted by excessive length. Significant effort to extract main points.
  3. 3Clear enough. Can understand the problem, approach, and results without undue effort. Core content clearly present: problem, method, findings, limitations.
  4. 4Well presented. Easy to follow, well-structured, appropriate level of detail. Target audience would get it quickly.
  5. 5Exceptionally clear. A pleasure to read. Complex ideas made accessible. Could serve as a model for how to present this type of work.

  1. The project makes a strong and specific argument: Africa's AI governance gap is less about writing policy and more about building the technical ability to verify deployed AI systems. The “Sovereign Safety Gap” framing is compelling, especially because the paper connects it to Nigeria’s mismatch between policy-capacity ranking and overall AI readiness, then broadens the claim to Kenya, South Africa, and Mauritius. The LAAI framework is also well motivated: using existing offensive-security and red-teaming talent for localized AI audits feels practical and regionally grounded rather than abstract.

    The strongest part of the paper is the healthcare pilot matrix. It shows the author understands that African AI safety failures may appear through local disease prevalence, underrepresented patient groups, and robustness failures from ambiguous or manipulated inputs. That makes the proposal more concrete than a generic “AI governance capacity” argument.

    The main weakness is that the paper stops at scoping. No actual audit is conducted, so we cannot tell whether LAAI would uncover failures that existing model evaluations miss, whether regulators would act on the findings, or whether offensive-security methods transfer cleanly to AI evaluation. The regional evidence is also thinner than the argument suggests: Nigeria has the strongest policy-capacity evidence, while the other country comparisons rely mainly on overall readiness rankings.

    To strengthen this, the next version should run one small audit on a real or open healthcare AI system, document actual failure cases, and show how findings would be handed to a regulator or institution. It should also explain more precisely which pen-testing practices map well to AI auditing and which require new methods.

    Overall, this is a thoughtful and well-written governance proposal with a strong theory of change, but it needs real audit evidence to move from persuasive framework to demonstrated contribution.

    Read full reviewShow less
  2. This paper presents a clear and compelling policy contribution by framing Africa's primary AI governance challenge as a "Sovereign Safety Gap" between policy and the technical capacity to verify deployed AI systems. The proposal to leverage existing offensive security expertise for adversarial AI auditing is practical and well motivated, while the four-stage auditing framework provides a structured approach for operationalizing AI governance. The Nigeria case study effectively illustrates the core argument.

    The main limitation is that the work remains a conceptual framework without validation through a real-world audit. The healthcare test matrix would be stronger with more Africa-specific evaluation scenarios, and the regional analysis would benefit from comparable pillar-level evidence across additional countries. A pilot evaluation of a deployed healthcare AI system and engagement with regulators would significantly strengthen the paper's practical impact.

    Overall, this is a thoughtful and well-argued position paper that offers a valuable framework for advancing AI governance through technical assurance rather than policy alone.

    Read full reviewShow less
  3. The thesis is genuinely interesting and well-argued: Africa's realistic governance leverage is technical verification capacity, not compute, and the continent's existing base of offensive-security practitioners can be redirected toward AI auditing. The disposition is honest throughout — the pilot is explicitly scoped rather than executed, the dual-use risk is raised unprompted, and the regional claim is hedged exactly where the data thins. But this is a position paper, not research. The "results" are an existing index's four overall ranks (Kenya 65th through Nigeria 72nd) reproduced rather than analysed, and a three-row test matrix that is a template, never run against a real system. The central empirical claim — that policy capacity outpaces verification capacity at a regional level — rests entirely on one country's pillar breakdown (Nigeria's 35th in policy versus 72nd overall); the other three have only an overall rank, so the regional pattern is asserted, as the paper itself concedes. The path forward is the one it names: execute the healthcare audit against a single deployed triage tool, and the methodology stops being a claim about what an audit could find and becomes one about what it does.

    Read full reviewShow less

Cite this project

@misc{jatto2026closing,
  title = {{Closing the Sovereign Safety Gap: A Localized Adversarial Auditing Framework for African AI Governance}},
  author = {Omonivie Cynthia Jatto},
  year = {2026},
  month = jun,
  note = {Submitted to Global South AI Safety Hackathon, an Apart Research Sprint},
  howpublished = {\url{https://apartresearch.com/sprints/projects/closing-the-sovereign-safety-gap-a-localized-adversarial-auditing-framework-for-african-ai-governance-6fbf}},
  url = {https://apartresearch.com/sprints/projects/closing-the-sovereign-safety-gap-a-localized-adversarial-auditing-framework-for-african-ai-governance-6fbf}
}

Build something like this at the next Sprint

AI Collusion Research Sprint · Oct 23 - 25, 2026