UK Officials Tested AI Coding Aides from Microsoft, GitHub, and Google - Observations Revealed
The UK government has recently concluded a significant trial involving the use of artificial intelligence (AI) coding assistants from Microsoft, GitHub Copilot, and Google Gemini Code Assist. The trial, which spanned across 50 different departments and involved over 1,000 tech workers, aimed to explore the potential benefits and challenges of integrating AI in government software development.
Nigel Douglas, head of developer relations at software supply chain security firm Cloudsmith, has voiced concerns about potential security issues with AI coding assistants. He emphasised the need for security-aware tooling or policy enforcement to prevent these assistants from unknowingly introducing vulnerabilities into critical government software ecosystems. Douglas also highlighted the importance of being able to respond to newly emerging threats that may impact what has already been deployed.
Despite these concerns, the trial revealed some positive outcomes. More than half (58%) of participants said they would prefer not to return to working without these solutions. Nearly two-thirds (65%) of users reported completing tasks more quickly, and 56% said they could solve problems more efficiently. Additionally, developers using these AI assistants are saving around one hour each day, equivalent to around 28 working days a year.
However, it's important to note that approximately 85% of government AI-generated code still requires manual editing by engineers. Martin Reynolds, the field CTO at software delivery platform Harness, has expressed doubts about the government's AI plans, questioning if they go far enough. He also pointed out that not much evidence of secure-by-design thinking has been observed in the use of AI coding assistants.
Google Cloud is thrilled about the positive impact its AI coding tool had on government workers during the trial. The study from Fastly found that developers often need to manually remediate faulty code, which negates the time savings provided by the technology. Despite this, 72% of users agreed that the AI assistants offered good value for their organization. Only 15% of code generated by AI assistants was used without any edits.
The use of AI coding tools by developers in certain countries is booming, with those developers being the most frequent users. The quality of AI-generated code has become a recurring talking point in recent months. As the UK government continues to explore the use of AI in its software development, it will be interesting to see how these concerns and benefits shape the future of AI integration in government tech.
