UK Government explores responsible AI adoption through new advisory panel

The Government Digital Service has established a Responsible AI Advisory Panel to guide the use of artificial intelligence across public services

The Responsible AI Advisory Panel is a pilot initiative that brings together expertise from government, industry, academia, and civil society, including frontline public sector experience.

It is chaired by Jeni Tennison, Executive Director at Connected by Data. The Responsible AI Advisory Panel was first introduced in March 2026 and has since held multiple meetings to shape early thinking on the responsible adoption of AI in government services.

The work aligns with other digital transformation priorities within the UK government, including efforts led by the Department for Science, Innovation and Technology.

Defining responsible AI in public services

Rather than relying on a single definition, the panel treats responsible AI as a broad set of principles that ensure technology is used safely and effectively in the public interest. These include service effectiveness, fairness, security, resilience, environmental sustainability, democratic legitimacy, and value for money.

A main concern is avoiding harms arising from poorly designed or poorly governed AI systems. In the public sector, failures can have wide-reaching consequences, from reduced trust in government services to direct negative impacts on individuals who rely on automated decisions or advice.

Early findings from government AI use

On the positive side, AI is already being explored to improve service delivery, reduce waiting times, and make public services more accessible. This includes support for people with additional needs, such as those who speak English as a second language or require assistance outside standard working hours.

At the same time, the Responsible AI Advisory Panel has identified significant difficulties in evaluating large language models. Systems similar to ChatGPT, Gemini, or Claude can produce fluent and confident responses, but testing their accuracy and reliability remains complex. This challenge is amplified in government settings, where reliable datasets of correct answers are often limited.

The panel also noted that while private sector applications may sometimes rely on disclaimers, public sector services require a higher standard. Incorrect or misleading AI-generated guidance can directly harm users and undermine trust in government institutions.

Key areas of focus

The panel’s work is structured around three main areas. First, it provides advice on individual AI projects and services to help ensure responsible design from the outset. Second, it offers strategic guidance to ministers on topics such as procurement, transparency, and digital sovereignty. Third, it supports efforts to spread responsible AI practices across government departments.

One example of emerging work includes engagement with AI-enabled tools such as Gov Voice, a reusable capability designed to improve efficiency and accessibility in public services. The panel is also reviewing early-stage AI applications in education, including personalised tutoring systems.

The Algorithmic Transparency Recording Standard is another key reference point, helping ensure that government departments document and disclose how algorithmic systems are used.

Limitations and future direction

The panel is advisory rather than a governance body, meaning it cannot certify whether AI systems are fully responsible. It also acknowledges that no single group can represent all perspectives or provide complete expertise across such a fast-moving field.

Over the coming months, the Responsible AI Advisory Panel will continue to test its approach through engagement with real government projects. Its work will conclude with an independent report summarising findings and recommendations.