The three primary cases — China, Rwanda, and Somalia — are all African or African-adjacent contexts. A framework that claims generalizability must demonstrate applicability beyond a single region. Vanuatu offers the most demanding test: it is geographically remote from Africa, has a distinct colonial history (Anglo-French Condominium), possesses a unique community governance tradition (kastom), and exhibits a pattern of cultural wealth combined with extreme state capacity limitation that mirrors the Tier 3 diagnostic without replicating the specific conditions of Somali state collapse.
Kastom (Bislama for "custom") is not a museum artifact — it is a living legal and social framework governing land tenure, conflict resolution, social hierarchy, ceremony, and the transmission of cultural knowledge. Chiefs (nakamal) are its custodians, deriving authority from knowledge of ancestral lore rather than state appointment. In many rural areas, kastom law operates parallel to and often supersedes national law. Kastom encodes ecological knowledge, oral narrative traditions, and intergenerational cultural transmission protocols that no external AI system can replicate or validate without community authorization.
The Somali Xeer and Vanuatu's kastom are structurally analogous: both are pre-colonial community governance systems that mediate conflict, transmit collective norms, and operate independently of formal state infrastructure. The decisive difference: Somalia's cadaverous state destroyed the Xeer's institutional conditions; Vanuatu's low-capacity state has largely left kastom intact. This creates a fundamentally different Tier 3 intervention logic. The goal in Vanuatu is not to rebuild community governance — it is to ensure that AI cultural tools report to kastom rather than bypass it.
This distinction refines the ICG framework's Tier 3 category. Somalia represents the cadaverous state: community governance destroyed alongside state capacity. Vanuatu represents a different Tier 3 configuration: state incapacity with community governance intact. Both require offline-first AI tools and community data sovereignty. But the intervention logic differs: Somalia requires rebuilding kastom-equivalent structures; Vanuatu requires protecting kastom from being bypassed by well-intentioned but externally governed AI projects. The TDCGP's community validation requirement (Req. 4) is the key mechanism in both cases, but for different reasons.
The Vanuatu Cultural Centre has collaborated with international linguists to document indigenous languages. Resulting archival materials are often held by external institutions. Communities whose oral traditions were recorded frequently lack meaningful access to the digital archives, let alone governance rights over how they are used, processed by AI, or commercially licensed. Kastom knowledge encoded in oral narrative is extracted without provision of data sovereignty — the Somalia dynamic in a Pacific setting.
With 138 languages averaging 1,760 speakers each, external AI teams cannot possibly possess the community-internal cultural knowledge required to validate annotation, transcription accuracy, or thematic organization. Which kastom stories are sacred and restricted? Which oral performances can be shared publicly and which are restricted to ceremony? These determinations require kastom authority, not linguistic expertise. Current AI documentation projects have no formal mechanism for kastom validation.
Most of Vanuatu's cultural heritage exists in remote island communities with intermittent electricity and no reliable internet. Cloud-based AI heritage systems are inaccessible by design in these contexts. The communities with the most endangered languages are precisely those furthest from infrastructure. This is not an implementation challenge — it is a structural incompatibility between current AI project designs and the communities those projects nominally serve.
Administrative execution capacity: minimal — no national AI cultural infrastructure, no state capacity for standardized Periodic Reporting. Legal infrastructure: formal ratification without enforcement capacity; kastom law operates as de facto primary legal framework in rural areas. Community participation mechanisms: high through kastom — but these mechanisms are bypassed, not incorporated, by current international AI heritage projects.
| Dimension | China (Tier 1) | Rwanda (Tier 2) | Somalia (Tier 3a) | Vanuatu (Tier 3b) |
|---|---|---|---|---|
| State Capacity | High — vertically integrated bureaucracy | Mid — rebuilding post-genocide | Cadaverous — all frames collapsed | Low — geographic dispersal limits reach |
| Community Governance | Functional, integrated with state | Rebuilding; Gacaca as instrument | Xeer destroyed by state collapse + commodity capitalism | Kastom intact — primary legal framework in rural areas |
| Cultural Wealth | Extensive; 100,000+ registered items | Rich; Kinyarwanda as administrative + cultural carrier | Extraordinary oral tradition; no policy capacity | Highest linguistic density per capita globally; sand drawing (UNESCO ICH) |
| AI Governance Model | State-directed; professional institutions | Hybrid; cultural AI embedded in admin capacity-building | Community-controlled; rebuild governance substrate first | Community-controlled; kastom councils as primary validators |
| Bidirectionality Status | Achieved — state provides public goods via cultural infrastructure | Achieved — Kinyarwanda NLP provides civic standing + admin efficiency | Absent — AI extracts without providing; Xeer destroyed | Absent — AI extracts without kastom governance rights |
| TDCGP Priority Requirement | Req. 4 — Community validation standards for professional institutions | Req. 2 — Data sovereignty embedded in admin programs | Reqs. 1+2+3 — ICA first; rebuild governance; offline-first | Reqs. 2+3+4 — Kastom sovereignty; offline access; kastom validation |
The three-tier structure applies directly to the Pacific context without modification. The capacity assessment dimensions (administrative, legal, community participation) generate the same Tier 3 classification for Vanuatu as for Somalia, though through different pathways.
The case reveals an internal distinction within Tier 3: cadaverous state (community governance destroyed) vs. low-capacity state (community governance intact). This distinction matters for intervention design but does not alter the TDCGP's four requirements — it only affects the sequencing and emphasis.
All four TDCGP requirements apply without modification to the Vanuatu context. The kastom council plays the same governance role as the community-level governance structures the TDCGP's Requirement 2 specifies. The framework requires no Pacific-specific adaptation.