Comment on No Siri AI in EU by Horst

@Michael Although I’m usually a fan of European regulation, in this case I find Apple’s reasoning pretty convincing.

Apple integrates Apple Intelligence as operating system service. The preferred execution environment is the local device. If necessary, the operating system decides whether to fall back to the cloud. Through Private Cloud Compute, Apple guarantees that even when data is processed in the cloud, no one other than the user has access to the processed data.

Apple can only provide this guarantee if it has complete control over the execution environment of the respective AI models.
For Gemini, they obviously have corresponding contracts with Google. That is likely to come at a cost.

Granting the same status to arbitrary models (in external cloud environments) would mean that the user completely loses control over access to their data. Due to its tight integration into the operating system, this is in no way comparable to an app.

One should also consider who is on the EU Commission’s side in this case… Every AI provider that would like to have unlimited access to the most private data of iOS users.

That doesn’t mean Apple hasn’t used similar arguments in the past to hinder the competition. But in this case, the reasoning seems sound to me.

Whether Apple will actually be able to run Apple Intelligence securely, however, is another matter entirely. Given how deeply integrated it is into the operating system, the notorious vulnerability of LLMs to attacks involving malicious input data strikes me as highly problematic.

Maybe it’s not such a bad thing that I have to wait and see how this plays out in the rest of the world before jumping in here in Germany.