The Cabinet Division has directed government officials to exercise strict caution when using AI tools for official work, warning that careless use could expose sensitive and confidential information to third parties.
In a notice issued recently, the division said the use of AI tools in government offices is increasing across various official activities. While these tools can improve efficiency and productivity, their careless use "could pose risks to the security and confidentiality of official information."
Officials have been asked to exercise the highest level of caution and responsibility when using AI tools and to follow existing laws, rules, regulations, and directives governing the protection and confidentiality of government information.
Why the warning now
The notice comes as generative AI applications capable of drafting, summarising, translating, and analysing documents become widely accessible. These tools are being used for an expanding range of tasks in government offices, which increases the risk of staff inadvertently feeding confidential material into external systems.
When officials paste documents or data into AI platforms, that information can be transmitted to third-party servers, where it may be stored, processed, or used to train future models. The distinction between routine document handling and potentially exposed data is not always clear to users who are not trained on the risks.
The directive creates a baseline expectation for public servants to assess what a tool will do with their input before using it. For those seeking structured guidance on using AI safely in governance contexts, the AI Learning Path for Policy Makers covers the practical boundaries of applying AI tools to official work.
What the directive requires
The division's notice did not ban AI tools, but it places the burden on officials to exercise judgment. Officials must confirm that any tool they use complies with the rules that already govern handling of confidential and sensitive information.
This includes verifying the security posture of third-party AI services, avoiding the insertion of classified content into open platforms, and ensuring that any automated processing occurs within approved and permitted systems.
The risk is not limited to security agencies or classified files. Routine government correspondence, procurement records, personnel files, and policy drafts all contain information that, if released, could compromise operations or the privacy of individuals.
For officials and public administration teams who want to stay current on how government bodies are approaching these risks, AI for Government compiles the latest guidance and use cases for AI adoption with appropriate safeguards.
Why this matters for government employees
Using a freely available AI tool for a routine task is not necessarily a violation for officials. But disuse to the environment through which government data travels - and how to verify who receives it - has now become an express part of the official's professional responsibility. This means that the government employees, it is helpful to method such protocols into daily workflow best practice before beginning to use the AI system.
Your membership also unlocks:
