Sometimes German Courts can be factually wrong while being substantively correct, and that turns out to be pretty good news for open social networks. Welcome to another update in the ongoing saga of ‘EU governance of open social networks is messy’, and to the surprise of everyone (read: me) we’re having an edition where a court ruling actually lives up to the strength of open social networks. Even more, the German Court explicitly quoted Mastodon and Bluesky as positive examples.
A quick TL;DR: Earlier this year, the Court of Justice of the European Union ruled on how to determine if a company is liable for content that is spread via it’s algorithms. It basically told the national courts in Europe: if you have a case that relates to a company that has distributed content via an algorithm and other people are mad about it, then here’s a test you should apply to determine who is responsible for what.
What’s at stake here is whether a service loses its hosting protection, the legal shield that lets you host other people’s posts without automatically becoming legally responsible for what they say. This sounds very dry and bureaucratic, and it absolutely is, but these kinds of tests and ruling are very influential, as it strongly impacts things like ‘What is Big Tech actually legally responsible for’. The problem with the test that the CJEU created is that it is written in a way that does not account for the complexities of open social networks well. If the company that hosts the data, controls the algorithm that distributes it, and is the only way that users can access this data, the test makes sense. But open social networks split this up, your data can be hosted for A, the algorithm operated by B, while the app on your phone is made by C, which makes it very unclear about who is actually responsible according to the ruling the the CJEU.
(I wrote a detailed analysis about this a few months ago, The Algorithm, Singular, if you’re a sicko who thinks that reading about the legal aspects of internet governance is fun (shout-out to the sickos))
Anyway, that has now happened: on 16 September, a court in Frankfurt ruled against Meta over scam content impersonating a financial education platform. Meta argued that it should benefit from hosting protection, the rules that limit a platform’s responsibility for content uploaded by its users. The court applied the CJEU’s test and said that Meta’s advertising auctions and feed algorithms gave it control over distribution. Not knowing about a particular scam did not get it off the hook. Meta was ordered to stop the infringements and provide information about them, and was found liable for damages.
So far, this is the part where the test works reasonably well: one company hosts the content and decides how to distribute it. But the court then addresses a rather important question. Does this interpretation mean that social networks in general lose their hosting protection?
Apparently chronological
The court explicitly says no, and to explain why, it points to Mastodon and Bluesky. On pages 33–34 of the judgment, it distinguishes Meta from providers that, in translation:
“essentially leave the selection of displayed content to their users”
It says this covers traditional online forums and would probably also apply to:
“purely chronological user feeds, as, to the court’s knowledge, Twitter originally and services such as Mastodon or Bluesky currently use”
The German word is used here is ‘gerichtsbekannt’, meaning ‘known to the court’. Apparently nobody in Frankfurt has found the Discover feed yet.
Bluesky has had custom feeds and a personalised Discover algorithm since 2023. So yes, the factual description is wrong. But Mastodon and Bluesky are being mentioned here for a specific reason: they are the court’s examples of how a social network can leave selection to users and still qualify for hosting protection. The court needs that distinction to explain why its ruling against Meta does not mean that every social network should be treated like Meta.
So even though the court does not actually understand how their own examples actually function, they did manage to land the underlying idea well, by treating user control over content selection as a reason to preserve hosting protection. Chronological feeds are just one way of doing that. Presumably, letting users choose their own feed algorithms is another.
The underlying logic of the court is solid here: On Mastodon, I can choose whom to follow and have their posts appear in order, which the court thinks is good. On Bluesky, I can do that too, but I can also choose a feed made by someone else, with a different way of selecting posts, or replace Bluesky’s own Discover feed. The fact that my chosen feed uses an algorithm does not make my choice disappear.
To be clear, the court has not actually decided whether choosing an algorithm satisfies that distinction, or how to assess Bluesky’s own recommendations alongside independent feeds. But for an open-network operator, this creates a useful argument: giving users control over their feeds could help preserve the hosting protection that makes running the service possible in the first place. That is considerably better news than simply being told that using algorithms might cost you your hosting protection.
In my previous article, I worried that courts would struggle to apply their platform-shaped rules to open social networks. Here, a court is explicitly recognising that social networks can organise control differently, and that this should matter for hosting protection, going as far as naming the networks as positive counter examples. Now we get to explain that the thing it likes about them can work with algorithms too.