AI on the sly? UK government stays silent on implementation

UK government officials have embraced AI technology at the highest levels of power, with thousands of civil servants using a proprietary chatbot called Redbox to generate draft briefings and analyze government documents. This previously undisclosed adoption of AI within the heart of government raises significant questions about transparency, accuracy, and the potential for biased outputs to influence policy decisions without public awareness or oversight.

The big picture: New Scientist has uncovered that at least 3,000 Cabinet Office staff who directly support Prime Minister Keir Starmer are actively using an in-house AI tool, with officials refusing to disclose how the technology is being implemented or what safeguards exist.

• Civil servants have conducted at least 30,000 interactions with the Redbox AI system, which can generate draft briefings and analyze government documents in seconds.
• Despite freedom of information requests to 20 government departments, officials have declined to release records of AI interactions, claiming such requests are “vexatious” and would require excessive resources to fulfill.

Why this matters: The integration of AI into core government functions represents a significant shift in how policy is developed, with limited public transparency about how machine-generated content might influence decision-making.

• One civil servant reportedly used the AI to synthesize 50 documents “in a matter of seconds” rather than spending a full day on the task.
• Large language models have well-documented issues with bias and accuracy, raising concerns about the quality of information potentially reaching senior officials.

Key details: The Department for Science, Innovation and Technology (DSIT), which oversees Redbox, has declined to answer specific questions about whether the prime minister or cabinet ministers are receiving AI-prepared advice.

• A second department, Business and Trade, acknowledged holding “over 13,000 prompts and responses” from staff interactions with Redbox.
• Officials have referred inquiries to DSIT rather than providing specific information about implementation practices or safeguards.

What they’re saying: Government representatives defend the AI implementation as a productivity enhancement that allows civil servants to focus on more important tasks.

• “No one should be spending time on something AI can do better and more quickly,” a DSIT spokesperson told New Scientist, adding that Redbox helps “harness the power of AI in a safe, secure, and practical way.”
• The spokesperson claimed the tool makes it “easier for officials to summarize documents, draft agendas and more,” ultimately speeding up work and freeing officials to focus on “shaping policy and improving services.”

Between the lines: The government’s refusal to disclose records of AI interactions suggests potential concerns about revealing how extensively machine-generated content may be influencing policy development.

• New Scientist’s investigation followed its earlier success in securing the “world-first release of ChatGPT logs under freedom of information legislation.”
• The classification of these information requests as “vexatious” raises questions about whether the government is using procedural barriers to avoid transparency.