AI Governance

Purpose

“With the economic promise and opportunity that AI brings, comes great social responsibility. Leaders across countries and sectors must collaborate to ensure it is ethically and responsibly developed, deployed and adopted.”

WEF, Jan 2024, AI Governance Alliance Briefing Paper Series

Open

Importance

• Responsibility: Define an AI-governance process to make sure who is responsible and where does the process sit

• Risk management: Mitigating data privacy, intellectual property and ethical risks associated with AI

• Business critical: AI Governance is critical to business value and efficiency and to limit enterprise risk

• Innovation: An improved organizational access to and adoption of AI systems, driving innovation and process efficiencies

The AI Governance framework should be evaluated use case by use case with a holistic perspective (start with the purpose of the use case and end with fact-checking of the output) and consider technical and procedural considerations for safe generative artificial intelligence (AI) models.

Key Pillars of the Unique AI Governance Framework

Open

A. AI Principles: guiding concepts and values

Goal: Privacy and Security

B. Data Access Control: Knowledge management and search

Goal: Accountability

C. Model Validation: Usage restriction to intended use cases/purpose

Goal: Transparency

D. Hallucination Management: Ensure high quality, unbiased outputs

Goal: Explainability

E. Standards and certifications: Set of practices and controls demonstrate compliance with laws

Goal: Reliability and Safety

AI Governance Committee in FSI

Unique is planning to host an AI Governance Committee with leading CDOs and compliance officers of our clients by the summer of 2024 to discuss key pillars of the AI Governance framework. We also planning to agree on major key elements of the AI Governance Framework and publish a Whitepaper about it.

Further helpful resources

Responsible AI in practiceUNDEFINED

FinanceGPT Etiquette: Do's & Don'ts

Data Leakage Prevention (DLP) - Shadow Documentation - Confluence (atlassian.net)

This page is still a work in progress and will be gradually adjusted and extended.