Firefox 148 Lets You Kill All AI Features in One Click (firefox.com) 48
Mozilla has released Firefox 148 for Windows, macOS and Linux, bringing a new AI Settings section that lets users disable all of the browser's AI-powered features in one click and then selectively re-enable the ones they actually want, such as the local translation tool that works locally rather than in the cloud.
The update also patches more than 50 security vulnerabilities -- none known to be under active exploitation -- over half of which Mozilla classifies as high risk, including five sandbox escape flaws and eight use-after-free bugs in the JavaScript engine that could allow code execution.
The update also patches more than 50 security vulnerabilities -- none known to be under active exploitation -- over half of which Mozilla classifies as high risk, including five sandbox escape flaws and eight use-after-free bugs in the JavaScript engine that could allow code execution.
dupe: kill with fire (Score:2, Insightful)
Re: (Score:2)
Re: (Score:2)
Fire in the hole!!!!!
A childbirth moment, if ever there was one.
inconsiderate? (Score:2)
Re: (Score:2, Interesting)
That's the catch. They use AI to report AI usage statistics. So when you turn the AI off nothing gets reported about your usage; ergo 100% of reports show AI usage.
Re: (Score:2)
"We've had no complaints"
Re: (Score:2)
What usage statistics? Their AI is local, except for the sidebar where you can login to some non-Mozilla providers.
Sure it does (Score:2, Informative)
Sure it turns off all the AI stuff, we totally believe you 100%.
Of course you'd spend billions creating this thing, forcing it into the software, then attempt to force everyone to use it, but then yeah of course you'd add a button to "turn it all off". Sure you would.
Well you have the source code... (Score:2, Troll)
Re: (Score:2)
Theoretically yes, but you'd have to go through a LOT of code to make sure there's also nothing to secretly turn it back on, or to later get an update to turn it back on, etc. At the end of the day a level of trust is required.
Re: (Score:1)
1) How many people would be good enough to spot the code?
2) We're also assuming that we really downloaded the actual source code. Maybe there are multiple versions masquerading under the same build number. Maybe there are A/B or Blue/Green tests going on with the downloads. How would we ever really know?
Re: (Score:2)
That's no problem, I'll just have my firefox AI agent...wait a minute!
Re: (Score:2)
Theoretically.
Re: (Score:2, Insightful)
Wouldn't it be possible to verify the kill switch is doing what it claims to?
IMO, you're missing the point.
Integrating features (ex. AI features) and then LATER adding controls to toggle them means it's an afterthought.
I made a test-first-development analogy on the last post of this, and it applies here as well. In development, most would agree that test driven development (you write the test for a feature first, then implement the feature until it passes all the tests) is an ideal way to do things, though people often skip to implementing first. Same here. In this case, they could
Re: (Score:3)
But what are you not trusting? Same as: if you don't want to print, don't call the Print menu item; now it is: if you don't want to talk to a chatbot, don't call the chatbot feature. That it's. The kill switch just removes them from your view. What do you think the local models could be doing in your back while you're not calling them?
Re: (Score:2)
Great example! When I print something, the printer drivers and the connection to the printer are not part of the browser. The browser has an interface to work with printing, not implementing it itself. A print menu item is fine, as would be a button or interface to make calls to an LLM for something. You must configure a printer before the browser can use it.
Why would you trust these new AI features? I've worked directly with LLM's. They go off the rails sometimes. They're not always right. Why would I give
Re: (Score:2)
Why would you trust these new AI features? I've worked directly with LLM's. They go off the rails sometimes. Why would I give it access to my personal data and the browser in which most of interfacing with the outside world gets done?
The answer to your question, "Why would you trust", is that they are not implemented as a multi-purpose chatbot waiting for prompts to perform general browsing actions. They're a single-purpose tool that takes a text in argument, and returns another text (for example, translation between fixed combination of languages, each of which uses a model which is downloaded individually at first usage). The resulting text is then displayed. That's all.
It's certainly additional attack surface for classical exploits,
Re: (Score:2)
One could wonder why browsers (and Firefox) include a pdf exporter, a pdf viewer, an svg renderer,...
PDF and SVG are file formats that can and are served on the web. Having support for parsing and viewing them does not seem at all abnormal to me.
In what way is an LLM needed? I think you're undercutting your point with these examples.
The reason to include the local model LLM features in Firefox is the same as with the PDF viewer: they're useful and people use them.
No, just stop with this broken analogy already. People need something with PDF support to view PDF documents. There are no documents that only a LLM can read. It's a false equivilence.
I grant you that they could spin off a number of features I quoted (all the non-html renderer) as add-ons that could even come suggested by default, or even included by default, but easy to remove. That's how the translation feature was offered initially. ...
YES. And that's what I'm suggesting should be done. This is a community driven product, after
Re: (Score:2)
In what way is an LLM needed?
Translation is necessary to read the web. Maybe not to you, but Mozilla addresses the needs of a vast world.
The new image description feature (implemented as Alt-text for PDF) is important to vision-impaired users of the web.
Re: (Score:2)
"Integrating features (ex. AI features) and then LATER adding controls to toggle them means it's an afterthought."
That's not how things were. The features are already opt-in, but some loud people wanted a kill switch. Let's go through features:
- AI sidebar: It's an iframe where you can load some commercial provider's site or a web-interface running on localhost. If you don't configure it to load something, it won't.
- AI summaries: When you trigger the first link preview you are asked if you want to download
Re: (Score:2)
The AI features are menu entries you can select, for example the page translation that uses a local model. They do nothing if you don't call them. The kill switch removes them from the menu. It's trivial to check the functions are not anymore in the menu after activating the kill switch.
Re: (Score:3)
Did Mozilla spend billions to add AI to Firefox? I doubt it.
The features they have added seem mostly mild, basically making it easier to hand off some data about what you are looking at to a 3rd party AI service of your choice. There is some in-browser stuff like translation, which has been there for a long time and is actually decent.
They seem to have done it right. The main issue is that it diverts engineering resources away from things that might actually turn Firefox's slow decline around.
Re: (Score:2)
No need for conspiracy theories when you have the source.
Re: (Score:2)
No need for conspiracy theories when you have the source.
I'd be surprised if 10% of the people on the planet have the skill to look through the source code and puzzle it out. And that's a super generous estimate.
Re: (Score:2)
The trick is not to read billions of lines, but the relevant diffs.
And in 2026 you can let AI summarize such things quite well.
Re: (Score:2)
The trick is not to read billions of lines, but the relevant diffs.
And in 2026 you can let AI summarize such things quite well.
So...trust the AI and the diffs? Who will review the AI's output, another AI?
All I'm saying is that using AI requires a lot more trust than it first seems like it does.
We can be pretty tech-savvy and still not understand what's going on under the hood. It could certainly still be running its AI junk stealthily in the background while being deliberately deceptive, masking what its doing just like malware does. I'm not sure we'd actually know without a lot of investigation.
Kind of like "Honey" and all the und
Re: (Score:2)
You can cut things short: If you distrust Mozilla, then the AI integration is not the most dangerous point you should examine. And Firefox has so many lines of code, that if you distrust Mozilla you should not use it at all. You may be able to trace their AI integration, but if you distrust you would need to suspect some backdoor that is in the core for 10 years. Have a lot of fun searching it.
Now add another button (Score:2)
One that will get rid of Sam Altman permanently.
Firefox is just an escape valve now (Score:5, Interesting)
I've pointed out again and again that Mozilla is too dependent on ublock origin while at the same time stuffing their browser with advertising instead of using their donation budget properly. I call them Mo$illa for a reason. As soon as Ladybird and Servo are out of beta most remaining Firefox users will dump Gecko for more web tech focused engines instead.
Mozilla is due for an Xfree86/Xlibre type rebellion, and I'm not afraid to say that because much of the open source world is changing with many new devs coming in because of Microsoft ruining Windows and with legacy code being rewritten faster then ever now. Mozilla is history now, Phoenix was 24 years ago, it's long due for another rebirth.
Re: (Score:2)
xlibre? You mean like telling in your software readme how the woke left suppresses you?
"Lets You Kill All AI Features" meh (Score:5, Insightful)
Already disabled (Score:3)
I only wish I could do more than disable: tear the code right out of firefox. I hate AI with a passion.
Despite the dupe, something good could come of it (Score:2)
If Mozilla really wanted to benefit the world, it would collect statistics on how many Firefox users deliberately disabled the AI features and publish them, or even aggressively market the results.
Re: (Score:2)
This exactly. I would LOVE to see that statistic.
And the next step in this evolution: if it's disabled in the browser, it should send a string to search engines, etc, to automatically disable AI in search results as well.
Step three: detect and warn about AI content on any pages you're visiting.
Need this to happen STAT.
Amazing. (Score:3)
Nice business plan
1: Add features that a privacy focused user base didn't ask for
2: Add feature to remove the feature
3: Get press
This is not true (Score:2)
Re: (Score:2)
That's because algorithms as simple as LRU are already "traditional machine learning".
And yet (Score:3)
You still can't kill the harassment message of a new update with one click.
Re: (Score:2)
Generate an enterprise policy for it: https://addons.mozilla.org/en-... [mozilla.org]
Best option, amongst bad ones (Score:2)
First off: Where in the hell did anybody get the idea that a web browser is for anything other than browsing the web???
OK, got that bit off my chest...
There's a serious disease infecting all the young coders in geekyland these days. It's name is AI. It seems that anybody who graduated with a computer related degree within the past decade or two now presumes [WRONGLY] that everything needs AI injected into it. It's like all the younger programmers are those stupid dancing syringe characters from Colbert's sh
Re: (Score:3)
Once AI is in a browser, is it safe to use the browser for things like banking?
Yes because Firefox does not implement anything problematic. Here what it does, using local models:
1) a button translate pages
2) a button to add descriptions alt-text to images in the pdf viewer (so that a screen reader for blind people can describe images)
3) guess a title for a tab group
4) summarise the first paragraph of a page into one sentence
https://support.mozilla.org/en... [mozilla.org]
It also adds a right-click option to query popular chatbots about the selected text (summarise, explain, quiz, improve text) and d
Re: (Score:2)
First off: Where in the hell did anybody get the idea that a web browser is for anything other than browsing the web???
Also, does anyone remember when web browsers were considered thin clients? I think the last time this was true was in the late 00s with netbooks. Then, more than a decade before the Al craze, browsers became these turbocharged Javascript engines that need multiple gigabytes of RAM to run.
zero clicks (Score:2)
I did one better: killed firefox AI with zero clicks.