I ran a poll on LinkedIn in recent days which asked “If a controller temporarily can’t access personal data on its systems because of the Crowdstrike/MSFT incident is it a personal data breach?”
I worded the question carefully.
50% of the 100-odd people who voted said “no” and 50% said “yes”. The latter group are wrong. I say this with some trepidation because there are people in that group whose opinion I greatly respect.
But here’s why they, and, indeed, the Information Commissioner’s Office and the European Data Protection Board, are wrong.
Article 4(12) of the GDPR/UK GDPR defines a “personal data breach”. This means that it is a thing in itself. And that is why I try always to use the full term, or abbreviate it, as I will here, to “PDB”.
This is about the law, and in law, words are important. To refer to a PDB as the single word “breach” is a potential cause of confusion, and both the ICO and the EDPB guidance are infected by and diminished by sloppy conflation of the terms “personal data breach” and “breach”. In English, at least, and in English law, the word “breach” will often be used to refer to a contravention of a legal obligation: a “breach of the law”. (And in information security terminology, a “breach” is generally used to refer to any sort of security breach.) But a “breach” is not coterminous with a “personal data breach”.
And a PDB is not a breach of the law: it is a neutral thing. It is also crucial to note that nowhere do the GDPR/UK GDPR say that there is an obligation on a person (whether controller or processor) not to experience a PDB, and nowhere do GDPR/UK GDPR create liability for failing to prevent one occurring. This does not mean that where a PDB has occurred because of an infringement of other provisions which do create obligations and do confer liability (primarily Article 5(1)(f) and Article 32) there is no potential liability. But not every PDB arises from an infringement of those provisions.
The Article 4(12) definition is “a breach of security leading to the accidental or unlawful destruction, loss, alteration, unauthorised disclosure of, or access to, personal data transmitted, stored or otherwise processed”. Let us break that down:
- A breach of security…
- leading to [one or more of]
- accidental or unlawful…
- 1. destruction of…
- 2. loss of…
- 3. alteration of…
- 4. unauthorised disclosure of…
- 5. unauthorised access to…
- personal data processed.
If an incident is not a breach of security, then it’s not a PDB. And if it is a breach of security but doesn’t involve personal data, it’s not a PDB. But even if it is a breach of security, and involves personal data, it’s only a PDB if one of the eventualities I’ve numbered 1 to 5 occurs.
Note that nowhere in 1 to 5 is there “unavailability of…” or “loss of access to…”.
Now, both the ICO, and the EDPB, read into the words “loss of…personal data…” the meaning, or potential meaning “loss of availability of personal data”. But in both cases they appear to do so in the context of saying, in terms, “loss of availability is Article 4(12) ‘loss’ because it can cause harm to data subjects”. I don’t dispute, and nor will many millions of people affected by the Crowdstrike incident, that unavailability of personal data can cause harm. But to me, “loss” means loss: I had something, and I no longer have it. I believe that that is how a judge in the England and Wales courts would read the plain words of Article 4(12), and decide that if the legislator had intended “loss” to mean something more than the plain meaning of “loss” – so that it included a meaning of “temporary lack of access to” – then the legislator would have said so.
Quite frankly, I believe the ICO and EDPB guidance are reading into the plain wording of the law a meaning which they would like to see, and they are straining that plain wording beyond what is permissible.
The reason, of course, that this has some importance is that Article 33 of the GDPR/UK GDPR provides that “in the case of” (note the neutral, “passive” language) a PDB, a controller must in general make a notification to the supervisory authority (which, in the UK, is the ICO), and Article 34 provides that where a PDB is likely to result in a high risk to the rights and freedoms of natural persons, those persons should be notified. If a PDB has not occurred, no obligation to make such notifications arises. That does not mean of course, that notifications cannot be made, through an exercise of discretion (let’s forget for the time being – because they silently resiled from the point – that the ICO once bizarrely and cruelly suggested that unnecessary Article 33 notifications might be a contravention of the GDPR accountability principle.)
It might well be that the actions or omissions leading to a PDB would constitute an infringement of Articles 5(1)(f) and 32, but if an incident does not meet the definition in Article 4(12), then it’s not a PDB, and no notification obligation arises. (Note that this is an analysis of the position under the GDPR/UK GDPR – I am not dealing with whether notification obligations to any other regulator arise.)
I can’t pretend I’m wholly comfortable saying to 50% of the data protection community, and to the ICO and EDPB, that they’re wrong on this point, but I’m comfortable that I have a good arguable position, and that it’s one that a judge would, on balance agree with.
If I’m right, maybe the legislator of the GDPR/UK GDPR missed something, and maybe availability issues should be contained within the Article 4(12) definition. If so, there’s nothing to stop both the UK and the EU legislators amending Article 4(12) accordingly. And if I’m wrong, there’s nothing to stop them amending it to make it more clear. In the UK, in particular, with a new, energised government, a new Minister for Data Protection, and a legislative agenda that will include bills dealing with data issues, this would be relatively straightforward. Let’s see.
And I would not criticise any controller which decided it was appropriate to make an Article 33 notification. It might, on balance, be the prudent thing for some affected controllers to do so. The 50/50 split on my poll indicates the level of uncertainty on the part of the profession. One also suspects that the ICO and the EU supervisory authorities might get a lot of precautionary notifications.
Heck, I’ll say it – if anyone wants to instruct me and my firm to advise, both on law and on legal strategy – we would of course be delighted to do so.
The views in this post (and indeed most posts on this blog) are my personal ones, and do not represent the views of any organisation I am involved with.
informationrightsandwrongs 