The Information Commissioner’s Office (ICO) has published guidance on AI-generated Freedom of Information (FOI) requests.

We have commented previously (here) on growing concern within Whitehall about the increased volume and complexity of FOI requests, and reports that the Government is considering changing the law – to the dismay of many news organisations – to introduce a lower threshold at which FOI requests can be rejected on costs grounds.

One of the factors identified as contributing to the increased time and cost of administering FOI requests is the use of AI to help draft them. This can not only lead to significantly more complex requests, but also to a greater incidence of legislation being misquoted or misinterpreted, or the need for substantial clarification before they can be processed.

While the guidance from the ICO is primarily directed at helping public authorities understand how they can confidently handle FOI requests that use AI, it is just as useful for those submitting the requests themselves so as to understand how they can use AI in a way that ensures that their requests are dealt with as efficiently and effectively as possible.

In particular, the guidance provides a template document for publication on a public authority’s website, setting out advice for requesters using AI, including the following:

• Only ask for information that you are genuinely seeking. The ICO notes that AI can sometimes generate broad or excessive wording that goes beyond the information that is needed.

• Ensure the request is clear, concise, and focused.

• Review the request to ensure that there are no factual inaccuracies. The ICO points to the risk of AI models misrepresenting rights under the Freedom of Information Act or referring to decision notices or cases that don’t exist. Authorities are encouraged to push back on such inaccuracies.

• Adopt an appropriate tone, as AI-generated content can sound abrupt or inappropriate.

The guidance makes clear that using AI will not make an FOI request invalid. However, public authorities may consider refusing to comply if, for example, a request is excessively burdensome, if AI is used to send repeated requests for substantially similar information or to pursue matters which are considered closed, or if a group of requesters use AI to send a high number of requests as part of a campaign.

To read the guidance in full, click here.