@EUCommission Addendum: It is highly doubtful whether Apple is capable of securely operating Apple Intelligence even within its own secure environment (on-device or in the Private Compute Cloud environment). The potential for attacks on the language models via the processed data is extensive.
h_albermann@mastodon.social
Indlæg
-
Have you heard of Apple's decision on the rollout of Siri AI in Europe? -
Have you heard of Apple's decision on the rollout of Siri AI in Europe?@EUCommission You should take a look at who supports this EU regulation. In this case, it’s anyone who wants full access to Apple users’ data.
And yes, Apple likes to use these arguments to shut out the competition. But in this case, things are a bit more complicated.
[3/3] -
Have you heard of Apple's decision on the rollout of Siri AI in Europe?@EUCommission As far as I know, Gemini as used by Apple is subject to precisely these restrictions in order to ensure data privacy. (https://security.apple.com/blog/expanding-pcc/)
If you now replace this secure AI environment with any other cloud provider, these guarantees disappear entirely. If Apple were to make the same APIs available to other AI models as it does to Gemini in the secure environment, users' data would be completely unprotected.
[2/3] -
Have you heard of Apple's decision on the rollout of Siri AI in Europe?@EUCommission I’m usually a fan of European regulations, but in this case, I find the Commission’s reasoning questionable.
Apple runs the AI models either locally on the device or in a specially secured cloud environment. For both, Apple guarantees that no one except the user has access to the processed data. The effort involved in operating the private cloud compute is enormous and well-documented.
In return, the AI then gains comprehensive access to user data.
[1/3]