Methodology and secondary AI review
Research covers 54 African UN member states. 9 complete candidate assessments received a fresh secondary AI review. 8 have an approved published rating. The other 46 countries remain unrated.
This is an AI-reviewed research release. The separate reviewer agents worked without the researcher’s scores until recording their blind assessments. This provides no human sign-off or external professional assurance.
What changed in version 1.1
On 17 October 2026, the project owner authorised a secondary independent AI reviewer for publication. The project’s G6 research-publication rule now accepts a fresh, independently contextualised AI review only after source verification, an immutable blind assessment, comparison with the research, remediation and an application check. Version 1.0 kept AI checks separate from publication sign-off. Country scope, dimension criteria, weights, confidence caps and gates G1–G5 remain unchanged.
Source dates
The evidence cutoff is 14 October 2026. Sources were rechecked on 18 October 2026, and release checks took place on 19 October 2026. Later access does not change the source cutoff. Unknown publication dates remain blank and are treated as limitations.
Scoring
| Dimension | Weight |
|---|---|
| D1 · Legal incorporation or endorsement | 30% |
| D2 · Scope and applicability | 15% |
| D3 · Effective status and transition maturity | 10% |
| D4 · Observed insurer implementation | 30% |
| D5 · Supervisory response and reconciliation | 15% |
All five dimensions are required. The weighted index is rounded to the nearest integer, with half points rounding up. Evidence gates can reduce the headline score. Confidence C caps the rating at 4 and is displayed as provisional. Confidence D has no published rating. Material unresolved conflicts cap a rating at 3. Missing evidence stays null and grey. A score of zero requires affirmative evidence.
Audited local insurer accounts, consolidated groups, statutory accounts and prudential returns are assessed as separate reporting layers. A parent group’s IFRS 17 accounts do not establish local subsidiary adoption. A purposive sample of audited insurers does not establish universal market compliance.
Fresh review results
| Country | Published assessment | Decision and coverage |
|---|---|---|
| Botswana | Unrated · provisional C | Retain provisional3/C but hold live headline grey. Verified local provisional amounts justify D4=2; current audited insurer and legal-scope/currency gaps prevent Ready publication. Official NBFIRA report and local Bona Life2024 report freshly read. BPOMAS audited medical-aid evidence retained in review files as adjacent scope; not counted as a licensed insurer audit. No market or subsidiary extrapolation. |
| Ghana | 4/5 | Publish candidate4/B after application check. D5 revised1→3 using the newer NIC-contributed official template evidence; all other dimensions accepted. SIC IAS28 qualification and low audited submissions remain visible. Fresh ICAG original adoption image and statutory-power scan; fresh SIC and Ghana Re audit/policy/company tables; fresh current IFAC scope corroboration; official2023 and2024 supervisory reviews. Two material entities provide a purposive implementation proxy, not life/non-life market representation. |
| Kenya | 4/5 | Publishable evidence supports proposed4/B after specified corrections and application check. Numerical proposal unchanged; local life separate-company corroboration added. Every decisive dimension source/claim traced. Fresh local opinion/policy/amounts and signature pages checked; historical nondecisive claims are explicitly sampled or held. No insurer census; group vs local statutory and prudential reporting layers remain distinct. Additional current group filings Britam2024/Jubilee2024/2025 independently read. Historical2022 readiness survey not freshly retrieved and does not drive D5. |
| Nigeria | 4/5 | Publish candidate4/B after application check. D2 revised4→5 for explicit broad insurer-type perimeter; D4 revised3→4 after fresh NSIA audit verification alongside AIICO; D5 revised1→3 after official annuity bridge and current SLA. Metadata and claim locators corrected. Fresh AIICO and NSIA local company audits including visual NSIA opinion/signature; fresh supervisor-host annuity circular; current May2026 regulator SLA; historical2016 Foundation profile and2022 prudential text checked for hierarchy/context.2023 unaudited/abridged filings remain supplementary. |
| Rwanda | 4/5 | Publishable evidence supports proposed4/B after specified corrections and application check. BlindD1 revised5→4; D2 rises4→5 from explicit current defined perimeter. Every decisive dimension source/claim traced. Fresh local opinion/policy/amounts and signature pages checked; historical nondecisive claims are explicitly sampled or held. No insurer census; group vs local statutory and prudential reporting layers remain distinct. |
| Seychelles | 3/5 · provisional C | Score3/C: operating-insurer consolidated evidence raises D4 to3, while unproved second mandatory cycle and full scope keep D2/D3 at4. Fresh circular and2023/2024 local consolidated operating-insurer audit examined. Audit is consolidated only. Historic law/regulation and official annual-report sections sampled; no national census, standalone subsidiary audit or second mandatory cycle assumed. |
| South Africa | 4/5 | Score4/B retained with corrected source/law/transition metadata and conservative statutory/supervisory boundaries. All decisive high-score sources independently reopened; local life/non-life insurer audits and selected group audits examined. Three local insurer examples are a credible proxy, not a quantified census. Inaccessible JSE/Old Mutual historical claims are retained outside score support. |
| United Republic of Tanzania | 4/5 | Publishable evidence supports proposed4/B after specified corrections and application check. BlindD1 revised5→4; original signed2024audit evidence accepted after local-render check. Every decisive dimension source/claim traced. Fresh local opinion/policy/amounts and signature pages checked; historical nondecisive claims are explicitly sampled or held. No insurer census; group vs local statutory and prudential reporting layers remain distinct. TAN-RE/Alliance primary-byte local renders resolve inconsistent web screenshots; EARE audit-placeholder row downgraded. |
| Zimbabwe | 4/5 · provisional C | Score4/C retained; active review4, narrow compliant audited implementation3, material adverse evidence retained. Fresh official article, IPEC report exact pages, original linked circular PDFs, Cell own audit, Fidelity qualification and domestic Old Mutual group audit examined. Narrow local implementation and significant adverse sector evidence remain visible. |
Earlier review provenance
The earlier review of remaining countries was reconstructed after temporary workspace files were lost. The original blind bytes for that batch are unavailable, so historical hash strings are not represented as verified seals. Those countries remain unrated. The fresh nine-country reviews have preserved blind artifacts and verified hashes.
Verification limits
Automated data, workbook and static interface checks were completed. Desktop and mobile browser interaction tests could not run in the available environment. The release record distinguishes these checks from browser verification. Remote source access can change after review.