HITL Governance

Version: 1.0 Effective Date: 1 March 2026 Owner: Altara AI Governance Office Classification: Internal Governance Policy Issued by: Altara AI · Part of the Navigate Compliance Group (Pty) Ltd

1. Purpose

This Human-in-the-Loop (“HITL“) Governance Policy establishes the governance principles, oversight mechanisms, and operational controls implemented by Altara AI to ensure that artificial intelligence systems remain subject to meaningful human oversight, accountability, ethical governance, and regulatory compliance.

Altara AI recognises that AI systems must augment — not replace — human judgment, particularly in environments involving governance, compliance, cybersecurity, fraud detection, risk assessments, investor protection, and regulatory decision-support.

This policy supports Altara AI’s commitment to:

• Responsible AI
• Human accountability
• Consumer protection
• Fairness and transparency
• AI governance best practices
• Digital trust and ethical innovation

2. Scope

This policy applies to:

• All AI-powered systems, tools, modules, and services developed, deployed, operated, or managed by Altara AI
• Internal AI systems used by employees, contractors, advisors, and partners
• AI-assisted outputs provided to clients
• AI-generated risk assessments, summaries, alerts, recommendations, classifications, or trust scores
• AI governance and monitoring platforms operated by Altara AI

This includes, but is not limited to:

• AI risk assessments
• AI trust scoring
• Fraud detection systems
• Scam analysis engines
• Regulatory intelligence tools
• AI governance dashboards
• AI document analysis
• Consumer protection monitoring systems

3. Governance Principles

Altara AI adopts the following HITL governance principles:

3.1 Human Accountability

Humans remain ultimately accountable for decisions, interpretations, actions, and outcomes generated using AI-assisted systems.

AI systems shall not independently make legally binding, regulatory, disciplinary, or high-impact decisions without human review and approval.

3.2 Human Oversight

AI outputs must be reviewable, challengeable, and overrideable by authorised human personnel.

Human reviewers must have sufficient authority, competency, and contextual understanding to validate AI-generated outputs.

3.3 Human-in-Command

Altara AI maintains governance structures ensuring that humans retain operational control over:

• AI deployment
• AI model configuration
• Escalation decisions
• Approval workflows
• Risk thresholds
• Monitoring controls
• Incident responses

3.4 Transparency

Users must be informed when they are interacting with or receiving outputs from AI-assisted systems.

AI-generated outputs must not be represented as definitive legal, regulatory, or professional advice unless independently reviewed and approved by appropriately qualified professionals.

3.5 Proportionality

The degree of human oversight shall be proportionate to:

• Risk level
• Regulatory impact
• Consumer impact
• Operational criticality
• Sensitivity of decisions
• Potential harm

Higher-risk use cases require enhanced human validation and governance controls.

4. Human Oversight Controls

Altara AI implements the following controls:

4.1 Review and Validation

AI-generated outputs may require:

• Human review
• Secondary approval
• Expert escalation
• Regulatory validation
• Compliance sign-off

depending on the risk classification of the activity.

4.2 Override Mechanisms

Authorised personnel may:

• Reject AI outputs
• Amend recommendations
• Escalate findings
• Suspend AI-driven processes
• Override automated classifications

where appropriate.

4.3 Escalation Procedures

Material risks, anomalies, or high-risk AI outputs must be escalated to designated governance personnel, including:

• Compliance officers
• AI governance leads
• Legal advisors
• Cybersecurity personnel
• Risk management teams

4.4 Auditability

Altara AI maintains audit trails where feasible, including:

• User interactions
• AI-generated outputs
• Human approvals
• Escalations
• Overrides
• Monitoring logs

to support accountability, transparency, and regulatory oversight.

5. AI Limitations and Disclaimers

AI systems operated by Altara AI:

• May produce inaccurate, incomplete, biased, or outdated outputs
• Operate probabilistically and not deterministically
• Should not be relied upon as the sole basis for decision-making
• Do not replace qualified legal, compliance, cybersecurity, financial, or regulatory professionals

Users remain responsible for independently validating outputs before reliance or implementation.

6. High-Risk Use Cases

Enhanced HITL governance applies to AI systems involving:

• Consumer protection
• Fraud detection
• Investor education
• Regulatory compliance
• Identity verification
• Trust scoring
• AI risk assessments
• Cybersecurity monitoring
• AI governance diagnostics

Such systems may require:

• Mandatory human validation
• Escalation workflows
• Governance approvals
• Enhanced monitoring
• Continuous assurance reviews

7. AI Governance Structures

Altara AI may establish governance structures including:

• AI Governance Councils
• Advisory Boards
• Ethics Committees
• Compliance Oversight Forums
• Model Risk Review Committees

to support responsible AI governance and oversight.

8. Monitoring and Continuous Improvement

Altara AI continuously monitors AI systems for:

• Performance degradation
• Hallucinations
• Bias indicators
• Emerging threats
• Adversarial manipulation
• False positives and false negatives
• Consumer harm risks

Governance controls may be updated periodically in response to:

• Regulatory developments
• Industry standards
• Emerging risks
• Operational findings
• Incident reviews

9. Regulatory Alignment

This policy is informed by evolving global AI governance principles and frameworks, including:

• OECD AI Principles
• NIST AI Risk Management Framework
• ISO/IEC AI governance standards
• EU AI Act principles
• Data protection and privacy laws
• Responsible AI best practices

10. Policy Review

This policy shall be reviewed periodically to ensure continued alignment with:

• Regulatory expectations
• Industry practices
• Operational realities
• Technological developments
• Organisational governance objectives

11. Contact

For questions relating to this policy or AI governance practices, contact:

Altara AI Digital Trust and AI Governance Part of the Navigate Compliance Group (Pty) Ltd Johannesburg, South Africa Website: www.altaracore.ai