Enzai: Compliance Assessment Tool

Case study from Enzai.

From:
Centre for Data Ethics and Innovation and Department for Science, Innovation and Technology
Published
12 December 2023
Use case:
Big data analytics, Data-driven profiling, Natural language processing and generation, Image recognition and video processing, Machine learning, Deep learning, Virtual agents or artificial conversational interfaces, Robotic process automation and decision management, Robotics and autonomous vehicles/systems
Principle:
Safety, security and robustness, Appropriate transparency and explainability, Fairness, Accountability and governance, Contestability and redress
Key function:
R&D, Product and service development, Human Resources, Marketing and sales, Customer services, Risk management, Strategy and corporate finance
AI Assurance Technique:
Data assurance, Compliance audit, Performance testing, Risk Assessment, Impact Assessment, Impact Evaluation, Conformity Assessment, Bias Audit
Assurance Technique Approach:
Technical, Procedural

Background & Description

With Enzai’s Compliance Assessment tool, organisations can translate their responsible AI policies into actionable insights, quickly and accurately measuring how their systems are performing against their policies. The tool consists of a range of different assessment techniques and formats, automated where possible to increase efficiency and reduce manual labour. To create safer and stronger AI systems, organisations can use the Compliance Assessment tool to standardise their approach to building and deploying AI.

With cutting-edge collaboration features specially designed for enterprises, the Compliance Assessment tool brings stakeholders from various areas of expertise together to review their AI. The granular system of controls and signoffs ensures that nothing is missed in the responsible implementation of AI.

Relevant Cross-Sectoral Regulatory Principles

Safety, Security and Robustness

The Compliance Assessment tool allows organisations to walk through an assessment step by step and ensure that every single requirement for the safety, security and robustness of an AI system has been addressed. A key feature of the tool are controls: a way of capturing evidence against certain requirements, providing a clear and concise way of demonstrating compliance. Users can define thresholds and benchmarks for their AI model performance metrics, including accuracy, fairness, robustness, and other key indicators of model quality.

Appropriate Transparency & Explainability

The AI Compliance Assessment tool is designed to make the process of establishing compliance for AI models and systems as transparent and clearly explainable as possible. For each AI model and system and organisation must choose which regulations and standards to check compliance against. The tool will then aid users in determining where they are in non-compliance, flagging those areas for transparency, and providing the appropriate steps to remedying the issue. Organisations can then choose with whom they would like to share this information to improve transparency - both internally and with their customers - to improve trust.

Fairness

The AI Compliance Assessment tool is where potential bias and fairness issues can be detected and resolved within an organisation’s AI systems. By assessing whether their AI models and systems are in full compliance with the relevant regulations and standards, organisations can know that their AI is not undermining fairness in any aspect.

Accountability and governance

Enzai’s AI Compliance Assessment tool is built to further accountability and governance with regards to an organisation’s AI governance practices. The tool enables organisations to check each of their AI models and systems against the relevant policies to ensure full compliance, with features which allow any problems to be flagged, tracked and resolved to ensure accountability.

Contestability and redress

The AI Compliance Assessment tool is the hub of contestability and redress on the Enzai platform. The Findings feature allows users of the tool to identify any issues found with an AI model or system’s compliance and flag it for review by other members of their organisation. The Findings can then be contested internally, and externally, and any problems found can be redressed,

Why we took this approach

Enzai strongly believes in the ability of standards and regulations to ensure that powerful technologies such as AI are used for good. The need for humans to oversee these incredible systems has never been greater and a strong governance structure is essential for oversight to be effective. The AI Compliance Assessment tool was developed to further this aim since it is essential for an organisation to be able to accurately and easily confirm the compliance status of each of their AI models and systems against relevant regulations and standards in order to ensure adequate oversight.

Benefits to the organisation using the technique

Enzai’s AI Compliance Assessment tool provides an organisation with an intuitive and seamless way to check the compliance of their AI models and systems against relevant policies and industry standards. This is the crucial step in ensuring full compliance with all necessary regulations and the highest possible governance standards. The AI Compliance Assessment tool provides greater visibility and accountability to an organisation’s AI usage and aids in increasing the robustness and effectiveness of governance and compliance activities.

Limitations of the approach

Enzai’s AI Compliance Assessment tool alone cannot ensure that an organisation meets its regulatory requirements. It is only one essential step of this process. Enzai has developed further tools (specifically, the AI Model Inventory, Policy Centre, and Governance Hub tools) which when used in conjunction with the Compliance Assessment tool allow an organisation to achieve the highest possible governance standards.

www.enz.ai

https://oecd.ai/en/catalogue/tools/enzai-ai-governance-platform

Further AI Assurance Information

Updates to this page

Published 12 December 2023
Contents