Copilot Extensibility Responsible AI Guidance

1. Summary

This guide to AI risk is designed to support NHS organisations in understanding and managing the complex landscape of AI risks, helping to ensure the responsible and secure deployment of AI technologies. It explains the different types of AI risks and poses key questions at each stage of development to guide users in assessing and addressing risks within their specific use cases. This document was written for the NHS.net Connect service in relation to Copilot Extensibility, any use beyond this intended purpose is at organisation discretion.

2. AI Risk Guidance

The AI Risk Guidance sections are designed to help assess your AI use case and identify potential risks within a given solution. They aim to ask the right questions at the right stages and provide guidance on the steps organisations should do to manage risks associated with AI.

It is split into 3 sections:

1. Initiation
2. Development
3. Post-Monitoring

Each section contains relevant questions to ask at their respective stage of a project. You can filter the areas that risks fall under, as well as whether a question is relevant to your use case. For example, some questions apply only to custom AI solutions where training data is used and fine tuning is involved.

3.AI Risk List

The AI Risk list provides a comprehensive view of the various types of risk you may encounter when developing AI systems, along with the broader categories and areas that they fall under. Each risk is explained in detail to help users gain a deeper understanding and assess whether it is relevant to their specific use case.

Risk teams can also use this section to map AI risks to any existing risks and controls they may already have in place.

Document Information and Downloads