| Framework Home | Cross-Pillar Governance | AISM | NEXUS |
Documentation updated: 2026-10-13. This guide clarifies the existing organizational score and explains two v1.0 companion methods. It does not change organizational scoring formulas or implement new CLI commands.
The October 2026 documentation update separates three questions that should be answered independently. The Sovereignty Score already existed. The Technology Contribution Profile (TCP) and Evidence Assurance Level are v1.0 additions, supported by a reusable Technology Card, validation guidance, and review examples.
| Instrument | What it assesses | When to use it | Method and record |
|---|---|---|---|
| AISM Sovereignty Score | Organizational maturity across five pillars and six dimensions | Baseline governance, prioritize organizational improvements, and reassess progress | Scoring methodology, self-assessment |
| AISM Technology Contribution Profile, v1.0 | What an exact product, technique, paper, or other artifact contributes to specific controls in a stated boundary | Review candidates, compare control fit, or define adoption conditions | TCP method and contribution rubric, Technology Card |
| AI SAFE² Evidence Assurance Level, v1.0 | How strongly each claimed contribution has been demonstrated | Judge whether evidence supports a claim and identify the next validation step | Evidence Assurance rubric, the card’s claim and control register |
The Technology Card is the record used for TCP and assurance review, not a fourth score. It is also distinct from the executable AISM Decision Card.
The full methods and Technology Card are maintained in this AISM directory. For a product or solution review, start with the technology evidence request. It provides a copyable provider request and checklist for collecting the artifacts needed to evaluate contribution and claim-level assurance.
not_assessed for unreviewed
entries; a maximum contribution does not describe complete pillar coverage.This is an illustrative review scenario, not a measured product result or a validated implementation. For a source-attributed design review, see the bounded ContractWarden card.
Suppose an organization identifies a gap in permission enforcement for one agent-to-tool path and considers a gate that checks each call against policy.
| Review step | What the assessor records | Permitted conclusion |
|---|---|---|
| Organizational baseline | Existing organizational evidence and the affected assessment cells | There is a governance gap; buying a gate does not resolve it by itself |
| TCP | Exact gate version, control IDs, interception point, policy ownership, coverage, bypass paths, outage behavior, and evidence | A design can describe a contribution; runtime enforcement requires supporting implementation and tests |
| Evidence Assurance | A reviewable design only, under the v1.0 rubric | E1 supports the design claim; it does not establish effective runtime blocking |
| Acceptance work | Positive and negative cases, unauthorized side effects, bypasses, revocation, restart, and legitimate task completion | A later E3 assessment requires a defined harness, reviewable results, and a bounded claim |
| Organizational reassessment | Local deployment, measured coverage, operators’ retained authority, and complementary controls | Reassess only the cells supported by the organization’s actual evidence |
See the validation guide for test planning and the incident evidence guide for failures and reconstruction. Neither guide claims these tests have been performed.
Report the organizational result with its methodology, scope, date, and assessor. Report technology contribution with its version, mapped claims, per-pillar coverage, evidence, and limits. Report assurance for each claim with the named AI SAFE² Evidence Assurance v1.0 rubric and independent-review attribution.
Do not multiply contribution by assurance, average technology scores into an organizational score, or treat assurance as certification. E5 operational evidence does not automatically establish E4 independent validation.
The organizational methodology’s Sovereignty Assurance metric, current CLI
E0-E5 numeric grade fields, claim-level Evidence Assurance, and
Challenge Lab C0-C5 have different meanings. The new methods change no CLI grade
weights, verification caps, schemas, or previous results. There is no implemented
TCP command or dashboard view. Use the Markdown card for this workflow.
A separate research note is optional for announcing these documentation changes. The linked methods already provide the rubrics, scope, and limitations. A future research note could explain the design rationale, alternatives, and validation plan. It should distinguish hypotheses from observed results and link the exact assessed revision. A completed study needs its own method, evidence, and review; publication alone does not validate these instruments.
| AISM Home | Technology Profile | Evidence Assurance | Technology Card | Nine review examples |
AI SAFE² v3.1 · Cyber Strategy Institute