Frameworks

Sector frameworks and the evidence model behind them.

Each framework defines its own controls, risk classes and evidence requirements. All four run on the same assessment infrastructure and the same four evidence tests.

Research programs

Different entry points. One evidence model behind all of them.

Each sector framework defines its own controls, risk classes and evidence requirements, and runs on the same assessment infrastructure.

Clinical AI · SaMD

AIGX Health

Clinical safety, evidence traceability, human oversight, privacy and lifecycle monitoring.

Banks · insurers

AIGX Finance

Model risk, explainability, fairness, third-party controls and regulatory reporting.

Autonomous agents

AIGX Agentic AI

Delegated authority, tool access, action boundaries, escalation and observability.

Public sector

AIGX Government

Public accountability, transparency, rights and safety, records and oversight.

Framework publications

Each framework is published as a title, with a reference and a review state.

A framework is not a position until it is written down, versioned and dated. These are the titles currently published or in preparation.

AIGX-RAIHFramework · v2.1

Responsible AI Governance: Healthcare Framework

Clinical safety, model validation, human oversight, monitoring, third-party controls and data privacy for one AI system in clinical or operational use, or in active pre-deployment.

AIGX-G-2026-01Framework · draft

Responsible AI Governance: Government Framework

Public accountability, transparency, rights and safety, records and oversight — set out as controls with named obligations, required artifacts and accountable owners.

Methodology

What makes published research defensible.

Criteria are published before findings are. Issuance is separated from remediation support. Every position carries a documented path to challenge it.

01

Published criteria

Scoring criteria, risk classes and evidence requirements are published before an assessment is scored against them.

02

Separated issuance

Platform workflow and remediation support are held apart from rating issuance and reviewer approval.

03

Documented appeal path

A rated organization can contest a finding through a recorded process, with the outcome traceable to the evidence.