Reference and glossary

Framework statistics

Item	Value
Version	1.3
Domains	17
Controls	358
Gate controls	71
Critical controls	70
High controls	241
Medium controls	43
Low controls	4

Domain reference

ID	Domain	Controls
A	Governance, Strategy and Accountability	11
B	Legal, Regulatory and Contractual Compliance	11
C	AI Use Case Intake, Approval and Risk Tiering	11
D	Data Governance, Lineage and Provenance	11
E	Data Security and Privacy Engineering	24
F	Secure Data Acquisition and Annotation	11
G	Model Development, Validation and Robustness	11
H	Prompt, Context and Retrieval Security	25
I	Inference, API and Runtime Security	25
J	Identity, Access and NHI Security	26
K	Agentic AI and Autonomous Action Governance	29
L	Third-Party, Model and Software Supply Chain Assurance	25
M	Monitoring, Detection and AI Security Operations	35
N	Human Oversight, Transparency and Impact Management	15
O	Resilience, Continuity and Recovery	36
P	Auditability, Evidence and Assurance	11
Q	Infrastructure, Platform and Environment Security	41

Badge dimensions

Dimension	Labels
Criticality	Critical, High, Medium, Low
Control type	Gate
Applicability	Mandatory, Triggered, Enhanced, Not applicable
Result	Unassessed, Partial, Supported, Assured, Gap, Gate fail, N/A

Criticality weights

Criticality	Weight
Critical	2.0
High	1.5
Medium	1.0
Low	0.75

Question answers

Answer	Meaning
Yes	Requirement is claimed as met
No	Requirement is not met
N/A	Proposed exclusion, subject to governance approval

Shared-responsibility roles

Code	Meaning
MP	Model Provider
OSP	Orchestrated Service Provider
AP	Application Provider
AIC	AI Customer
CSP	Cloud Service Provider
Shared	Responsibility divided between named parties
ND	Not determined

Assurance components

Component	Weight
Design effectiveness	15%
Implementation effectiveness	20%
Operating effectiveness	20%
Verified evidence	25%
Coverage	10%
Resilience	10%

Audit-readiness components

Component	Weight
Assurance	50%
Ownership clarity	15%
Verified evidence adjusted for coverage	35%

Assurance score

Score	Threshold
1	Gate fail or assurance at/below 25
2	Assurance above 25 and below 55
3	Assurance at least 55 and below 75
4	Assurance at least 75
5	Assurance at least 90 and audit readiness at least 85

Caps

Condition	Cap
Gate failure	25%
Critical control with evidence below 75	54%
Critical control without passing test	74%
High control with evidence below 50	54%
Failed, ineffective or exception test	25%

Result precedence

N/A
Unassessed
Gate fail
Gap
Assured
Supported
Partial

Evidence workflow

Requested
Received
Reviewed
Accepted
Rejected
Expired

Evidence quality

Poor
Fair
Good
Strong

Test results

Positive:

Passed
Pass
Effective

Adverse:

Failed
Fail
Ineffective
Exception

Other:

Pending
Not tested

Effectiveness labels

Effective
Partially effective
Ineffective
Not tested
Insufficient evidence

Residual-risk labels

Low
Moderate
Elevated
High
Critical
Unknown

N/A decision fields

Rationale.
Category.
Decision owner.
Independent approver.
Approval date.
Review date or reassessment trigger.
Compensating controls.
Evidence references.

Scope modes

Scope	Meaning
All GTSAF controls	Full canonical control population
Selected AI system	Per-system scope derived from a complete, conflict-free authoritative route

Scope calculations

Applicable controls

Controls factually relevant to the selected system. This is a breadth measure, not a risk, compliance or effectiveness score.

Applicability pending

Controls awaiting sufficient, conflict-free facts. Pending controls are not treated as passed or out of scope.

Out of scope

Conditional controls for which the complete route contains no factual trigger.

Baseline depth

Applicable controls requiring the ordinary assurance depth.

Depth-weighted workload

Planning measure calculated as Baseline × 1 + Triggered × 2 + Enhanced × 3.

Assurance intensity

Depth-weighted workload divided by Applicable controls × 3, expressed as a percentage.

Factual scope driver

A recorded system characteristic capable of activating one or more conditional controls. Driver counts may overlap.

AI Assist actions

Per control

Applicability analysis.
Evidence assessment.
Bounded test plan.
Effectiveness analysis.
Residual-risk analysis.
Recommendations.
Monitoring.
Cross-control dependencies.

Active scope

Executive assurance summary.
Material gaps.
Evidence and testing gaps.
Ownership issues.
Residual-risk decisions.
Monitoring needs.
Prioritised remediation roadmap.

Glossary

ACRS

Agentic Capability Risk Score. Product-based risk classification using dependency, action autonomy, access scope and harm potential.

Applicable

A control or requirement included in the selected assessment boundary.

Assurance

The degree of confidence supported by assessment answers, evidence, testing, ownership, coverage and resilience.

Assured

GTSAF result for a control with no No answers and assurance of at least 75%.

Audit readiness

The degree to which another reviewer can reproduce and defend the conclusion from the record.

Compensating control

An alternative safeguard reducing risk where the primary requirement is not fully met.

Conformance

Yes answers divided by all applicable question rows, including unanswered rows.

Control

A required governance, security or assurance outcome represented by a canonical GTSAF ID.

Control owner

The party accountable for implementing or operating the control.

Evidence owner

The party responsible for producing and maintaining evidence.

Enhanced

An in-scope control requiring deeper evidence and stronger testing because of elevated risk or exposure.

Finding

A governed record describing a control weakness or assurance exception.

Gate

A control whose failure can block a positive assurance conclusion.

Gap

A result produced by at least one applicable No answer.

Mandatory baseline

Applicable controls assigned the ordinary assurance depth. Baseline, Triggered and Enhanced are mutually exclusive and reconcile to the applicable-control population.

Narrative sufficiency

Quality of implementation and customer-responsibility explanation; separate from evidence.

Operating effectiveness

Whether a control works in practice over the assessed period.

Residual risk

Risk remaining after considering implemented controls and compensating measures.

Scope

The named system, components, environments, data, users, suppliers, period and exclusions covered by the assessment.

Supported

A positive but below-Assured result: at least one Yes, no No, and assurance below 75%.

Triggered

An applicable control requiring additional assurance because a factual system feature or exposure is present. ACRS or governance weighting may also deepen the assurance requirement without changing factual applicability.

Verified evidence

Linked, reviewed and quality-evaluated artefacts or operating records supporting the assessment.

Crosswalk disclaimer

GTSAF references to external frameworks provide traceability. They do not establish automatic compliance, exact equivalence or certification.