The government has recently released guidelines for software engineers working in government departments regarding the use of artificial intelligence (AI) based coding assistants.
The Government Digital Service (GDS) document titled AI coding assistants for developers in HMG highlights the potential risks associated with using AI coding assistants, especially when developing, maintaining, and deploying production services from a single environment.
GDS emphasizes the importance of following good practices in development platforms and deployment infrastructure to minimize risks when utilizing AI coding assistants. The guidance suggests that software engineering teams within government departments can mitigate risks by working transparently and implementing main branch protections.
In addition, the guidelines recommend maintaining strict separation and auditing of production secrets access, as well as incorporating multi-stage deployment with adequate test coverage and vulnerability scanning in software development pipelines.
Due to the unpredictable nature of AI coding assistants, GDS advises against relying on specific responses to prompts in source code and build pipelines unless extensive testing has been conducted to mitigate potential issues.
The release of this guidance follows a successful four-month trial involving over 1,000 software engineers utilizing AI to enhance productivity. The Department for Science, Innovation and Technology (DSIT) reported that AI assistants have the potential to save government software developers the equivalent of 28 working days per year.
By leveraging AI assistants, developers were able to build more software to support government-led digital initiatives, leading to increased efficiency. DSIT estimates that these AI assistants could help the government achieve £45 billion in savings by streamlining public sector operations.
During the trial, developers and engineers from 50 government departments tested AI coding assistants from Microsoft, GitHub Copilot, and Google’s Gemini Code Assist. The majority of users expressed satisfaction with the tools, acknowledging their value for organizational productivity.
Minister of Technology, Kanishka Narayan, commented on the positive results, stating that the use of AI by engineers demonstrates a desire to improve efficiency while ensuring accuracy in delivering public services.
