Over the last 12 months organisations have seen a huge increase in the number of individuals using AI to make Data Subject Access Requests (SARs). Although AI can help individuals understand and articulate their rights under the UK GDPR, it can also create challenges for organisations responsible for responding to these requests. It is now essential to understand how to manage and respond to AI assisted SARs effectively.
It is trite law that an individual does not need to draft an SAR personally for it to be valid. A request does not become invalid simply because it has been generated or assisted by AI, nor does the requester need to refer expressly to the UK GDPR, or use particular legal terminology.
The key consideration remains whether it is sufficiently clear that the individual is seeking access to their personal data. AI generated requests are often lengthy and unnecessarily complicated, requesting extensive categories of information including metadata, audit records, CCTV footage, call logs and information held across multiple systems. However, organisations should approach AI assisted SARs as they do for any other SAR, and cannot ignore it simply because it is AI generated and contains extensive and/or irrelevant legal references, repetitive wording or other inaccuracies.
Where a request is genuinely unclear, organisations can ask the individual to clarify what personal information they are seeking. This may be particularly helpful where the request covers a substantial volume of information or includes categories of data that are difficult to interpret, as is often the case with those produced with AI.
The ICO's guidance on the clarification of SARs can be found here. In short, if you intend to clarify the scope/nature of the request, you should do this without delay and write to the requester well before the end of the statutory response period of one-month. Further, where a request is particularly difficult to understand because it contains duplication or overlapping categories of data, rather than simply asking for clarity it can be helpful to set out your understanding of what is being requested and what (if any) elements of the request do not actually constitute a personal data. Please note that seeking clarification on the scope of a SAR, pauses the clock for the purposes of the statutory deadline, and re-starts on the day after the data subject provides this clarification. Seeking clarification can be particularly important where a data subject has used AI, which (as above) can result in an overly complex/convoluted request: it may prove to be the case that they are really only interested in a much narrower scope of information.
Depending on the nature of the request, it may be necessary to extend the response deadline by up to a further 2 months due to its complexity. However, any extensions should be considered on a case by case basis; and the requestor's response to any request will likely be relevant to this assessment.
Complexity might arise where information is difficult to retrieve, extensive third-party information requires consideration, or multiple data protection rights have been exercised simultaneously.
The fact that a request has been drafted using AI, or is particularly lengthy or confusing, does not automatically make it complex. Any extension must be justifiable and the requester must be informed of the extension and the reasons for it within the initial one-month period.
Organisations are required to undertake a reasonable and proportionate search for personal data falling within the scope of a SAR. This does not mean searching every database, backup system or electronic record regardless of whether relevant information may be held there.
The organisation must determine the appropriate scope of its searches, taking account of the nature of the request, the systems used and where the individual's personal data is likely to be located. Organisations should document their search methodology, including the systems and departments considered, the information included or excluded and the reasons for those decisions. This provides an important audit trail and demonstrates accountability, should the matter later be referred to the Information Commissioner's Office.
A comprehensive SAR drafted with the use of AI, may specifically request information that will be difficult to identify, or will be structured in a way that demands a “no stone unturned” approach. Again, an organisation must consider whether it is reasonable and proportionate to search for this data. The recent High Court case of Ashley v HMRC [2025] suggests that this assessment is not limited to the search for data. Citing the earlier Court of Appeal case of Dawson-Damer v Taylor Wessing LLP [2017], which previously held that the question as to whether disproportionate effort was required in responding to a SAR was not limited to the challenge of searching for the data but also complying with the request more broadly, the Court in Ashley v HMRC considered this "would include, for example, time spent addressing whether exemptions applied to any of the personal data identified and addressing the extent to which the data should be redacted". In practice, it may be that some categories of data are easy to search for but would require the extensive redaction of third party data before it can be disclosed. For example, where data which is responsive to a SAR, held in aggregated spreadsheets of employee data such as lists of employees' holiday/sickness records or performance metrics, could only be disclosed to a data subject after removing the data of all other employees.
When assessing which information to disclose in response to a SAR, an organisation should consider all of the relevant circumstances, including the type of data being requested, the likely importance of the information to the data subject and the feasibility of disclosing it. It is for the data controller to demonstrate that a search and disclosure exercise is unreasonable or disproportionate, and so it must give due consideration to each element of the SAR before opting not to provide the information sought.
Organisations may want to consider how they can leverage AI themselves to respond to SARs; with developments in this area being used to streamline the search, review and redaction process. Already there are a range of tools available which market themselves as being able to complete an entire first-level review of a SAR, without requiring a human to review a single document. This capability potentially extends the use of automation and bulk filtering which has been used for some time. These tools typically work through a prompt or series of prompts which can be used to categorise documents as responsive, or non-responsive to scope of the SAR. Some tools go further, having functionality to apply redactions and/or consider the application of exemptions under the UK GDPR, for example, by identifying whether a document may be legally privileged, or where its disclosure may prejudice any negotiations with the data subject.
The tools available have huge potential to improve the quality of a SAR response and reduce the substantial time and effort that can be required to undertake the review process. However, any organisations handling SARs with the aid of AI must ensure they are aware of their obligations and devise a process that is consistent with those obligations.
Further, whilst the application of some exemptions may be clear cut, others will require careful judgement and a detailed understanding of the surrounding circumstances - for example, the extent to which litigation privilege might apply to internal investigations into an underlying employment issue. Human review therefore remains essential when using AI, as documents may be incorrectly categorised, relevant personal data may be overlooked, responsive information may contain personal data relating to third parties, or exempt material may be inadvertently disclosed. Any such errors may expose an organisation to complaints, regulatory action or litigation, including from other individuals whose personal data might be share. This may be of particular concern when responding to a SAR in the context of a complaint or claim, where the careless disclosure of material could prejudice your position or serve to escalate matters in the context of a dispute. Any use of AI must be overseen by individuals with a detailed knowledge of the UK GDPR.
In the age of AI, organisations should have clear internal procedures for managing SARs, AI assisted or otherwise. Staff should receive appropriate training to recognise requests, identify potential inaccuracies, verify identity and understand when clarification or specialist advice is necessary.
Ultimately, AI assisted SARs should not be viewed as a new category of data protection request as the existing legal framework continues to apply. However, organisations that adopt a consistent, proportionate and well-documented approach will be much better placed to meet their statutory obligations protect individuals' rights and manage the additional challenges that AI generated SARs may present.
For more information about AI assisted SARs, please contact our data protection team.