In a recent interview with The Sunday Times, Deputy Prime Minister Oliver Dowden revealed the government’s intention to harness artificial intelligence (AI) to assist ministers in reducing their workload and streamlining the civil service. Dowden highlighted ongoing efforts to implement AI technology in the handling of ministerial red boxes, which contain official documents prepared by ministerial offices.
AI trial for ministerial red boxes
Deputy Prime Minister Dowden announced that the government is currently conducting a trial to assess the application of AI technology in processing ministerial red box submissions. These submissions play a crucial role in the decision-making process for ministers, involving the extraction of essential information from various sources and aligning it with parliamentary, judicial, and public sentiment contexts.
Dowden proposed the use of AI to handle the tasks of reading, summarizing, and compiling these important documents, potentially replacing the need for civil servants to perform these functions. The objective is to significantly reduce the time required for document processing and enhance the quality of decision-making by providing immediate access to information summaries.
The integration of AI into government processes aligns with the broader objective of downsizing the civil service workforce. In November 2022, Chancellor Rishi Sunak abandoned Prime Minister Boris Johnson’s plan to cut 91,000 civil service jobs, which accounted for approximately 20% of the workforce. However, in October 2023, former Health Secretary Jeremy Hunt unveiled new plans to reduce 63,000 civil service positions as part of cost-saving measures.
Safety concerns surrounding AI use
While the government explores the potential of AI in the civil service, it is important to address safety concerns associated with AI technology. The government’s own guidance on AI usage within the civil service cautions against over-reliance on large language model (LLM) systems, which are likely to be utilized for document processing.
The guidance emphasizes the risk of false information emerging at any point in AI-generated content and stresses the importance of cross-checking facts and assertions, irrespective of their apparent authority.
The government’s advice further acknowledges the susceptibility of generative AI to biases and misinformation. It recommends the need for thorough verification and appropriate citation of AI-generated content. Notably, the guidance explicitly discourages the use of AI for authoring messages and summarizing facts for communication to others, potentially conflicting with Deputy Prime Minister Dowden’s plans.
Warnings about publicly available AI tools
The government’s guidance extends its caution to the use of publicly available AI tools, including Google Bard and ChatGPT. These tools, accessible to the public, are subject to the same concerns related to bias, misinformation, and the need for verification.
It is important to note that the government established an Office for AI in 2018, primarily tasked with guiding the national AI strategy. However, this office’s role does not extend to acting as a steward of AI utilization within the government, leaving the responsibility for AI implementation and safety in the hands of individual departments.
