Handling AI-assisted SARs in the workplace
This website will offer limited functionality in this browser. We only support the recent versions of major browsers like Chrome, Firefox, Safari, and Edge.
What has changed?
We have seen a sharp increase in individuals using generative AI to draft subject access requests (SARs). The resulting requests can be lengthy and highly detailed, often listing numerous custodians, systems, search terms and categories of information. They may also be repetitive or unclear and include inaccurate legal references or requests for information which do not, in fact, fall within the right of access.
The ICO’s new guidance on AI-assisted SARs confirms that none of this changes the fundamental position: a SAR is not invalid simply because AI has been used to draft it. What matters is whether the individual is seeking access to their personal data, not the form of the request or the accuracy of the legal terminology used.
As our colleagues explain in Same right, new author: the ICO's guidance on AI-assisted SARs - Burges Salmon, the underlying right has not changed. The practical challenge for employers is working out what an increasingly lengthy and sophisticated-looking request actually requires them to do.
What does this mean for employers?
The starting point is not to treat the AI-generated request as a search protocol that must simply be followed line by line. Instead, employers should cut through the drafting and identify the substance of the request: what personal data is the individual actually seeking, over what period, in relation to which issues, and where is that information realistically likely to be held?
That distinction can be particularly important in an employment context. A SAR made alongside a grievance, disciplinary process, dismissal or Employment Tribunal claim might ask the employer to search every email, Teams message and document mentioning the employee over several years, across a lengthy list of custodians. It may also contain requests for explanations, documents or other information which fall outside the scope of the right of access. The fact that an AI tool has generated that list does not mean the employer must adopt it as its search methodology. The ICO expressly recognises that AI-assisted requests may encompass information outside the scope of a SAR.
Clarification can be part of the solution
Where the request is unclear, employers should consider engaging with the individual before embarking on extensive searches. The aim should be to understand what information the employee is actually trying to obtain and, where appropriate, agree a more focused and workable scope.
For example, rather than immediately searching all correspondence concerning an employee, it may be appropriate to establish whether their concern actually relates to a particular grievance, redundancy exercise or decision, and the relevant period and individuals involved. A constructive conversation about scope can benefit both sides: the individual may receive the information they actually want more efficiently, while the employer avoids carrying out extensive searches which add little or nothing to the response. The ICO confirms that organisations can seek clarification where a request is unclear and encourages engagement with requesters in appropriate circumstances.
In an employment dispute, we would generally recommend documenting that engagement. It may be appropriate to confirm any discussion about scope in writing, particularly where there is an ongoing grievance, disciplinary process or litigation.
What is a “reasonable and proportionate” search in practice?
This is perhaps the most important point for employers faced with an expansive AI-assisted SAR. The obligation is to undertake a reasonable and proportionate search for the individual's personal data, not an unlimited search for every piece of information that could conceivably fall within the request.
There is no single formula for determining what is reasonable and proportionate. Employers should make a considered assessment of the request and the information they hold. In practice, that is likely to involve looking at factors such as:
A requester's suggested custodians, systems or search terms may be useful, but they should not automatically dictate the search exercise.
The key is to be able to demonstrate why the searches undertaken were reasonable and proportionate in the circumstances. Employers should therefore keep an audit trail recording the scope adopted, any clarification obtained, the systems and custodians searched, relevant search parameters, and the reasons why particular searches were or were not undertaken.
A practical approach to an AI-assisted SAR
When a lengthy AI-assisted request lands with HR or the data protection team, we would suggest working through it in the following order:
The temptation when faced with a detailed, legalistic AI-generated SAR can be either to take every request at face value or to dismiss the drafting as AI-generated. The better response is to cut through the form of the request, identify the personal data the individual is actually seeking and put in place a search exercise that is targeted and proportionate.
For a fuller overview of the ICO’s guidance, see Same right, new author: the ICO's guidance on AI-assisted SARs - Burges Salmon.
Want more Burges Salmon content? Add us as a preferred source on Google to your favourites list for content and news you can trust.
Update your preferred sourcesBe sure to follow us on LinkedIn and stay up to date with all the latest from Burges Salmon.
Follow us