The Central Digital and Data Office (CDDO) has unveiled a new Generative AI Framework, designed to equip civil servants with tools to navigate this technology safely and securely.
The framework provides practical considerations for anyone in government planning or developing a Generative AI solution, and expands on the initial guidance provided by CDDO in June 2023.
It is based on a set of ten principles focusing primarily on large language models (LLMs), as these have received the most attention, and have the greatest level of immediate application in government.
These ten principles are: understanding what Generative AI is, ethical implications, tool security, quality assurance controls, managing the full Generative AI lifecycle, identifying use-cases, cross-government collaboration, transparency, skills building and governance structures.
CDDO said the framework is not intended to be a detailed technical manual due to the fact that the field of Generative AI is still developing rapidly, with best practice in many areas having not yet emerged.
"It is necessarily incomplete and dynamic," said David Knott, Chief Technology Officer for Government. "We will update it frequently as we learn more from the experience of using generative AI across government, industry and society."
He added: "It is intended to be accessible and useful to non-technical readers as well as to technical experts. However, as our body of knowledge and experience grows, we will add deeper dive sections to share patterns, techniques and emerging best practice (for example prompt engineering)."
The framework was created in collaboration with experts from government departments, arm’s length bodies, other public sector organisations, academic institutions and industry partners.