907
Firefox 148 introduces the promised AI kill switch for people who aren't into LLMs
(www.xda-developers.com)
this post was submitted on 25 Feb 2026
907 points (98.9% liked)
Technology
81907 readers
4534 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
@Feyd@programming.dev @technology@lemmy.world
I'm not referring only to the feature per se, I'm also referring to any pop-up designed to appear throughout the navigation to "remind the user about the superb features".
Said pop-up is explicitly mentioned on their "confirmation dialog" upon turning off (screenshot attached below):
It speaks volumes about how much a dark pattern this is, the fact that the opt-off has a confirmation dialog, while the further proceeding with logging in with Anthropic/OpenAI/Google/Meta account doesn't seem to have a confirmation dialog.
And the fact that the confirmation feels "menacing" and defaulted to cancelling the opting-off (i.e. pressing "esc" or clicking outside the window; one must click the primary-colored "block" button which, contrasted to a grayish "Cancel" button, may psychologically induce the user into thinking "block" is a dangerous action), quite similar to the
about:configwarning screen.Ah, and the clanker options: notice the lack of alternative options for those who want a custom clanker, such as DeepSeek, Qwen, Z AI, Brazilian Maritaca IA and Amazônia IA (to mention some non-Chinese LLMs), or even something running locally through ollama. Seemingly no option for using a custom, possibly self-hosted LLM endpoint. The fact that all the options offered are all heavily corporate options (with Mistral being the "least corporate" of them all, but still Global Northern nonetheless) might tell us something...
All of these dark patterns, among others not mentioned, are the object of my critique, not just the fact that Mozilla is shoving clankers unto Firefox.
Whenever a feature needs an invasive pop-up and the opt-out brings up a second pop-up that requires further confirmation (but none seems to be offered upon actually using said feature), it is called a dark pattern, no matter if said feature requires further configuration.
@dsilverz @technology @Feyd
A modal in general is not defined as a dark pattern (not sure why you say that).
And in this case a modal is used to manage a user journey "subtask", which is a request to confirm a potential disruptive action: users may use firerox AI features for long, before deciding to turn them off, deeply changing their experience with the product.
I agree that it could have been done as a full page, but it is fine also as a modal on desktop (not mobile viewport)
@skamu@mastodon.uno @technology@lemmy.world @Feyd@programming.dev
Maybe I'm overly idealistic when it comes to software but, IMHO, a software (especially a browser) should be the least distractive possible. My point about modals was about feature announcement pop-ups ("Now you can do Y... Click on Z menu to get into Y"), the ones which Mozilla Firefox explicitly mentioned within the confirmation dialog, as well as the said confirmation dialog which, as far as I could find about, is one-sided, for there are no confirmation dialog to the other action, which is to activate the clankers.
The ideal workflow, to me, is as follows: the user launches the browser software, the main UI opens minimalistically listing the most frequently accessed websites and the pinned bookmarks, the user clicks on some shortcut or types in some URL, then the browser fetches the network content from said website, parses it, fetches whatever else needs to be fetched for the specific website, renders it visible on the screen, then let the user interact with the page as they please, without a MS Clippy-like behavior of reminding the user "It looks like this page has links, you can summarize them using a clanker" on a frequent basis.
Lynx, for example, is the perfect example of this, it's not an utopia I'm imagining: I type
lynxand I press enter, then Lynx executes and brings its TUI, then I press g and type the URL of a website, and it fetches and does what needs to be done in order to bring up the website to the TUI. No cluttered interface except for the short list of keyboard shortcuts at the bottom which don't require user interaction nor disturb the UX. That's KISS approach.When a browser has a MS Clippy-like behavior and, most importantly, when a browser brings potentially unwanted features turned on by default, whose opt-out requires the user to go through some sort of gymnastics while the usage of said feature is asymmetrically easy (seemingly no "confirm you want to use the clanker? The clanker may have access to the following: page content, currently open tabs, credentials on the page, etc..." like the opt-out confirmation dialog lists exhaustively about "enhancements that will be unavailable while the user opts out of Firefox AI enhancements"), again: perhaps I'm being too pedantic but, to me, it smells, it looks, it behaves and it whispers like a dark pattern.
@dsilverz @technology @Feyd
Hi Deamon, don't get me wrong, modals in general are bad and it is a "last resource". However, I agree with the solution of Firefox's UX designer, asking for a confirmation before turning on/off the AI functionalities (as it is a disruptive action that affect overall users XD). We may argue that maybe it needed to be done via a full page... ? Or maybe not using a "switch" in the first place. Anyway, all good. It is nice to see people with this kind of concerns 👍
@skamu@mastodon.uno @technology@lemmy.world @Feyd@programming.dev
The thing is, there doesn't seem to be confirmation before turning clankers on (at least I didn't find screenshots in this regard), but there is such a confirmation before turning the whole thing off (that is, from the default-on state Mozilla pushed unto the software upon updating/installing).
If both situations involved double confirmation dialog in a symmetrical manner ("are you sure you want to proceed with activating this feature?" coexisting with "are you sure you want to opt-off from this feature?"), that would be fair. Pretty annoying, but fair. But this fairness doesn't seem to be happening, no confirmation dialog seems to exist for actually using the feature. The only thing similar to a "confirmation" during further usage of "AI Enhancements" would be the authentication step from whatever clanker was chosen from the suspiciously-biased list of clankers (ChatGPT, Claude, Gemini, Llama, Mistral; no non-Western options such as Qwen or DeepSeek, for example).
How disruptive would be turning off a feature that is far from being essential to browsing (and, in practice, may end up rendering the whole browsing experience worse with inaccurate summarization and potential vulnerabilities (prompt injection, remote code execution, etc), produced by pieces of software explicitly labeled as "it may produce inaccuracies")?
Not to say how, as I mentioned initially, the entire premise of bringing it as default-on with now the added "right" to "opt-off" is, itself, non-consensual relationship, insofar the user didn't seek it by themselves. Clankers would be a nice feature for some niches and use cases (again: I myself use LLMs, but it stems from my own decision to do so, not because it was pushed onto me; something I opted-in), but it should be voluntarily sought, installed and turned on by the user as they please, not as "default-on" option.
Sure, no problems, that's reciprocal, we're good! Throughout my exchanges in this entire thread, I tried to keep it respectful (at least when it comes to the debate and my peers; of course I'm fiercely criticizing Mozilla Corporation, because they were once the ones who "will never sell your data") and trying to debate the idea and not the peer's person.
My concerns, in the end of the day, are just an attempt to advocate for the total, non-negotiable autonomy and Free Will (as far as Free Will can get in a deterministic cosmic existence) of users, far from just my own; and this involves denouncing potential corporate biases whenever a corporation brings up another brick in the already-tall wall of enshittification, naming and shaming corporations for their greedy corporate behavior.