MODEL RISK & VALIDATION STANDARD

OVERVIEW

Analytical models may assist AIMM in collecting, organising, analysing and presenting Commercial Intelligence.

Models may include proprietary methodologies, internally developed tools, third-party software, Artificial Intelligence systems, machine learning technologies, statistical techniques or other analytical processes.

No analytical model is infallible. AIMM recognises that all models have inherent limitations and that model outputs should be considered within the context of AIMM's overall governance framework.

This Standard establishes general governance principles and does not constitute a contractual commitment regarding any specific model, methodology or validation process.

OBJECTIVES

This Standard is intended to support:

• responsible model governance;

• proportionate model validation;

• operational resilience;

• appropriate human oversight;

• continuous improvement;

• protection of proprietary Commercial Intelligence;

• responsible management of model risk;

• sound decision-support processes.

SCOPE

This Standard may apply, where appropriate, to:

• Artificial Intelligence systems;

• machine learning models;

• statistical models;

• analytical algorithms;

• data processing routines;

• screening methodologies;

• classification models;

• workflow automation tools;

• internally developed software;

• third-party analytical technologies.

The application of this Standard remains at AIMM's discretion.

MODEL GOVERNANCE PRINCIPLES

Fitness for Purpose

Where AIMM considers it appropriate, analytical models may be selected or developed for their intended operational purpose.

The suitability of any model may vary according to the nature of the Commercial Intelligence activity being undertaken.

Human Oversight

Analytical models are intended to support, rather than replace, AIMM's governance processes and commercial judgement.

Where AIMM considers it appropriate, model outputs may be reviewed before being incorporated into Commercial Intelligence Deliverables.

Proportionality

The level of governance applied to any model may reflect its intended use, complexity, operational importance and perceived risk.

Not all models require identical governance processes.

Continuous Improvement

Models, methodologies and analytical techniques may be modified, refined, replaced or discontinued at any time where AIMM considers such changes appropriate.

MODEL RISK

Model risk may arise from factors including:

• limitations within underlying data;

• changes to source information;

• assumptions inherent within analytical methodologies;

• software defects;

• external system changes;

• technological developments;

• inappropriate application of a model;

• limitations of Artificial Intelligence systems.

AIMM seeks to manage model risk through governance processes proportionate to the nature of the activity being undertaken.

VALIDATION APPROACH

Where AIMM considers validation appropriate, it may include one or more of the following activities:

• methodological review;

• logical consistency review;

• comparison against alternative information sources;

• testing using representative information;

• reasonableness assessment;

• operational review;

• governance review;

• periodic reassessment.

The extent and methodology of any validation remain at AIMM's discretion.

MODEL PERFORMANCE

The performance of analytical models may be reviewed periodically where AIMM considers such review appropriate.

Performance indicators may include factors such as:

• operational usefulness;

• consistency of outputs;

• suitability for intended use;

• identified limitations;

• changing business requirements.

No model is guaranteed to produce accurate, complete or current information in every circumstance.

THIRD-PARTY MODELS

AIMM may utilise third-party software, Artificial Intelligence technologies or analytical tools as part of its operations.

AIMM does not control the design, operation, maintenance or continued availability of third-party models.

Accordingly, AIMM makes no representation or warranty regarding the performance, availability or future development of third-party technologies.

MODEL CHANGE MANAGEMENT

Analytical models, software tools and methodologies may be updated, modified, replaced or withdrawn by AIMM at any time.

Changes may be made to reflect:

• technological developments;

• business requirements;

• operational experience;

• recognised governance practices;

• legal or regulatory developments;

• improvements in analytical capability.

AIMM is under no obligation to notify Clients or other parties of such changes.

CLIENT RESPONSIBILITIES

Commercial Intelligence Deliverables are provided in accordance with the applicable contractual documentation.

Recipients remain solely responsible for:

• independently reviewing Commercial Intelligence Deliverables;

• conducting their own legal, commercial, technical, financial and compliance due diligence;

• determining whether information is suitable for their intended purposes;

• making their own commercial decisions.

Nothing in this Standard transfers those responsibilities from the Recipient to AIMM.

CONFIDENTIALITY & PROPRIETARY METHODOLOGIES

The analytical models, validation processes, algorithms, methodologies and governance techniques used by AIMM may constitute confidential information and proprietary intellectual property.

Nothing in this Standard grants any right to inspect, access, reproduce, reverse engineer or otherwise use AIMM's proprietary methodologies except where expressly permitted by applicable contractual documentation.

RELATIONSHIP WITH OTHER POLICIES

This Model Risk & Validation Standard should be read together with:

• AI Governance Framework

• Supplier AI Assessment Standard

• Information Security Policy

• Data Retention Policy

• Compliance

• Commercial Intelligence Framework

• Commercial Intelligence Licence

• Terms of Business

• Privacy Policy

• Website Terms of Use

• Acceptable Use Policy

IMPORTANT NOTICE

This Model Risk & Validation Standard provides a general description of AIMM's internal governance approach to analytical models, Artificial Intelligence systems, software tools and related methodologies.

The performance of model governance, model validation, testing, monitoring or review activities by AIMM does not constitute a representation, warranty, certification or guarantee regarding the accuracy, completeness, timeliness, reliability or suitability of any Commercial Intelligence Deliverable or any information derived from analytical models.

Recipients remain solely and exclusively responsible for conducting their own independent legal, commercial, financial, technical, operational, regulatory and compliance due diligence before relying upon or acting upon any Commercial Intelligence Deliverable or any information contained within it.

AIMM's analytical models, Artificial Intelligence systems and validation processes are decision-support tools only. They are not intended to replace the Recipient's own review, verification, risk assessment or independent decision-making processes.

Recipients are solely responsible for determining whether any Commercial Intelligence Deliverable or related information is accurate, complete, suitable or appropriate for their own purposes.

AIMM accepts no responsibility for any commercial decisions, transactions, actions, omissions, losses, liabilities, damages or consequences arising from or connected with the use of, reliance upon or interpretation of any model output, analytical assessment, Commercial Intelligence Deliverable or related information.