Wired Op-Ed Pushes Opt-In as Default for Generative AI

The pattern is by now familiar enough to earn a weary shrug. You open an app you have used for years, and a new AI feature is already on. If you want it off, there is a toggle, usually a few menus deep, that you can go and find. The Wired opinion piece making the rounds this week is a plea to stop calling that consent.

The author writes they are sick of opt-out toggles for automatically enabled generative AI features, and argues it is past time to make opt in the default for sensitive ones. That is a small semantic switch and a very large behavioural one. Companies know that almost nobody changes a default, and so offering an opt-out becomes a way of taking data while sounding as though they asked. The setting that lets platforms train on user content is switched on when the user arrives, and only switched off if they go and find it.

The timing is not accidental. It lands the same week Meta pulled its Muse Image tool, which, as Wired reported, had automatically opted public Instagram profiles into being fodder for generative AI remixes that anyone could tag into a prompt to generate images using someone else's likeness. After public criticism and a SAG-AFTRA call for members to opt out to protect their likeness, Meta said the feature 'missed the mark' and withdrew it.

The honest caveat is that the piece is a short opinion column, not a policy proposal, and what I can see of it does not name a full list of offending product surfaces or draw a sharp line between AI features you interact with and background training-data collection. Those are different consent questions and deserve different toggles.

What is worth watching is which platforms move to opt-in on their own, before a regulator or the next viral union statement does it for them. Mozilla has already made 'AI features are fully opt-in' part of Firefox's pitch, and the more the default-on approach costs Meta-sized companies in trust and rollbacks, the more that stance starts to look less like a niche privacy flex and more like the safer default for anyone shipping consumer AI.