Privacy theatre, Apple edition

We’re being slowly conditioned to accept complete loss of privacy. Ubiquitous security cameras become Meta glasses on every face, the tracking tech in ankle monitors turns into Smart watches, targeted CSAM scanning (think of the children!) turns into scanning everything. Every surveillance technology slowly morphs into something with better UX that we willingly adopt.

As mentioned previously, we already live in a world where you have to assume your emails will be fed to ChatGPT, Gemini, or some other “assistant”.

And now, OpenAI has just announced that ChatGPT for Mac can read all your iMessages. This will certainly be adopted by many, and I’m sure other platforms will follow if there’s no pushback.

This is particularly perverse for a few reasons:

• iMessage is end-to-end encrypted. There’s an expectation of privacy because of this (Apple’s marketing leans on this). But your correspondent can now upload messages, before encryption or after decryption, to a large language model; with no indication of this on your side (granted, they could copy-paste it before, but now it’s automated).
• Because ChatGPT gets full access, it also gets full access to archives going back as far as they’re stored. Your private messages from 2 years ago? They might be heading for OpenAI’s servers right now.
• Anything that ends up on OpenAI’s infrastructure is fair game for US authorities thanks to the CLOUD Act. They just need to ask.

If you’re an organisation that uses iMessages internally, particularly if that includes sensitive information, it might be time to switch to a secure alternative like Signal or Threema.

And even if that’s not your profile, it’s worth thinking about what you type from now on.

Colin