This section is where the theoretical rubber meets the empirical road, making it the perfect place to insert Mitchell’s diagram. > We should use Mitchell's visualization as the central connective tissue of this section. The narrative should guide the reader through the diagram to explicitly show:

• What we want vs. What we have: Use the diagram to visually contrast our ideal theoretical framework (from Section 4) with the variables currently available in the ILIA 2025/2026 data.

To maximize its impact and read as an airtight executive proposal, we should rewrite the narrative introducing this framework so it explicitly ties back to our core theoretical pillars (3.1 to 3.4). Instead of presenting the table and principles as a standalone idea

This section is incredibly strong because it materializes the sovereignty debate into tangible infrastructure. However, it plunges directly into "Hawkins et al." without framing the concept of the "technical stack" first. Before citing the authors, let's open with a conceptual anchor that explains why a layered approach is necessary. We should introduce the idea that AI sovereignty cannot be measured as a single, flat variable because technological dependency operates across a deeply asymmetric, multi-layered hardware and software stack.

Let's introduce this tension first. By explaining that possessing the physical or legal capacity to control AI does not automatically guarantee sovereignty or ethical alignment, the reader will better appreciate:

• Why Roberts (2024) warns against the "measurement trap" of rewarding raw control.
• Why Santaniello (2025) speaks of sovereignty as a "discursive instrument" that can mask deep dependencies.
• Why Lehuedé’s (2024) distinction between extractive logics and grassroots projects is so vital for a Latin American context.

This conceptual entry point will make the closing question: "sovereignty for whom, over what, and at whose cost?".

This section does a great job of showing the scale of the research, but it currently reads too much like an internal developer log or a repository README. We need to elevate the tone to make it look like a polished academic/institutional methodology.

Let's refine it based on the following: * Highlight the LLM Pipeline professionally: Keep the mention of the LLM-assisted synthesis workflow used to extract the 147 structured units from the 88-source corpus. This shows advanced and systematic methodological innovation. * Remove internal technical noise: Strip out operational details that don't add academic value, specifically the SSL error, the manual data-entry workaround for the 2026 chapter, the .yaml configuration files, and the local Excel file paths (e.g., data/ILIA 2025/...). * Focus on the Integration: Keep the core logic of how the 88 papers were prioritized (the 5 lens criteria) and how they map against the ILIA 2025/2026 excel files to ensure measurement continuity, but frame it conceptually rather than referencing specific repository snapshots.

Great introduction. This is exactly where we establish the "WHY" or the the core justification for analyzing data sovereignty and its vital importance to the ILIA's evolution. To strengthen it further, let's explicitly outline our contributions here. We need to detail how adding this new Data Sovereignty section will serve as a major contribution to the ILIA, specifically by providing the region with actionable indicators to measure effective agency and strategic interdependence, rather than just tracking isolated AI assets.

To fix this and make the chart cleaner, we should shorten the dimension names in the visualization code. For example: * Legitimacy, rights, and sustainability --> Legitimacy & Rights * Institutional and regulatory agency --> Institutional * AgencyCompute and cloud agency --> Compute & * CloudInnovation, and application agency --> Innovation & AppsData and knowledge agency --> Data & Knowledge

To make it immediately actionable for ILIA, let's explicitly analyze it using 4 distinct quadrants: * High-High (Top-Right): True strategic agency. Countries that have successfully coupled infrastructure capacity (3.3) with institutional steering power (3.2). * Low-Compute / High-Agency (Top-Left): "Regulatory-first" posture. Strong data protection and frameworks, but severe infrastructure bottlenecks. This group perfectly justifies our recommendation for regional pooling (3.4). * High-Compute / Low-Agency (Bottom-Right): The "measurement trap" warned about in 3.2. Physical infrastructure presence without the regulatory leverage to audit or steer it. * Low-Low (Bottom-Left): Structural development traps (3.4), lacking both technical stack capabilities and regulatory oversight.

Excellent conceptual foundation. This section is critical for shifting the mindset from theoretical sovereignty to practical application. However, to make this argument truly impactful, we should introduce and anchor the core concepts first before diving into the specific literature.

Before quoting Couture, Pohle, or Repetto, let’s add a brief introductory setup that explicitly defines:

• What "Autarky" means in the context of AI (the illusion of absolute digital isolation/self-sufficiency).
• What "Relational Sovereignty" implies (agency and negotiation within an interdependent global network).

By defining this contrast upfront, the reader will immediately understand why the proliferation of multiple "sovereigns" (Repetto 2025) and Barasa’s "continuum of strategic postures" matter. It will also create a much stronger baseline for the final argument: why full-stack self-sufficiency is unrealistic for Latin America and the Caribbean (LAC), and why the region must focus on strategic leverage instead.