5 ms·

> Large language models are something else entirely*. They are black boxes. You cannot audit them. You cannot truly understand what they do with your data. You

> Large language models are something else entirely*. They are black boxes. You cannot audit them. You cannot truly understand what they do with your data. You cannot verify their behaviour. And Mozilla wants to put them at the heart of the browser and that doesn't sit well. Am I being overly critical here or is this kind of a silly position to have right after talking about how neural machine translation is okay? Many of Firefox's LLM features like summarization afaik are powered by local models (hell even Chrome has local model options). It's weird to say neural translation is not a black box but LLMs are somehow black boxes that we cannot hope to understand what they do with the data, especially when viewed a bit fuzzily LLMs are scaled up versions of an architecture that was originally used for neural translation. Neural translation also has unverifiable behavior in the same sense. I could interpret some of the data talk as talking about non local models but this very much seems like a more general criticism of LLMs as a whole when talking about Firefox features. Moreover, some of the critiques like verifiability of outputs and unlimited scope still don't make sense in this context. Browser LLM features except for explicitly AI browsers like Comet have so far had some scoping to their behavior, either in very narrow scopes like translation or summarization. The broadest scope I can think of is the side panels that show up which allow you to ask about a web page with context. Even then, I do not see what is inherently problematic about such scoping since the output behavior is confined to the side panel.

Cheer2171 10mo ago
No, it is disqualifyingly clueless. The author defends one neural network, one bag of effectively-opaque floats that get blended together with WASM to produce non-deterministic outputs which are injected into the DOM (translation), then righteously crusades against other bags of floats (LLMs). From this point of view, uBlock Origin is also effectively un-auditable. Or your point about them maybe imagining AI as non-local proprietary models might be the only thing that makes this make sense. I think even technical people are being suckered by the marketing that "AI" === ChatGPT/Claude/Gemini style cloud-hosted proprietary models connected to chat UIs.
- koolala 10mo ago
  I'm ok with Translation because it's best solved with AI. I'm not ok with it when Firefox "uses AI to read your open tabs" to do things that don't even need an AI based solution.
  - kbelder 10mo ago
    There's levels of this, though, more than two: local, open model local, proprietary model remote, open model (are there these?) remote, proprietary model There is almost no harm in a local, open model. Conversely, a remote, proprietary model should always require opting in with clear disclaimers. It needs to be proportional.
    - koolala 10mo ago
      The harm to me is the implementation is terrible - local or not (assuming no AI based telemetry). If their answer is AI then it pretty much means they won't make a non-AI solution. Today I just got my first stupid AI tab grouping in Firefox that makes zero intuitive sense. I just want grouping not from an AI reading my tabs. It should just be based on where my tabs were opened from. I also tried Waterfox today because of this post and while I'd prefer horizontal grouping atleast their implementation isn't stupid. Language translation is a opaque complex process. Tabs being grouped from other tabs is not good when opaque and unpredictable and does not need AI.
    - enriquto 10mo ago
      What do you mean by "open"? Open weights, or open training data? These are very different things.
      - kbelder 10mo ago
        That is a good point, and I think the takeaway is that there are lots of degrees of freedom here. Open training data would be better, of course, but open weights is still better than completely hidden.
user3939382 10mo ago
Firefox should look like Librewolf first of all, Librewolf shouldn’t have to exist. Mozilla’s privacy stuff is marketing bullshit just like Apple. It shouldn’t be doing ANYTHING that isn’t local only unless it’s explicitly opt in or user UI action oriented. The LLM part is absurd bc the entire overton window is in the wrong place.
- PunchyHamster 10mo ago
  It's frankly desperate trend chasing from management that lost after starting from near total market domination, and have no idea what to do now.
  - takluyver 10mo ago
    > starting from near total market domination That's not really accurate: Firefox peaked somewhere around 30% market share back when IE was dominant, and then Chrome took over the top spot within a few years of launching. FWIW, I think there's just no good move for Mozilla. They're competing against 3 of the biggest companies in the world who can cross-subsidise browser development as a loss-leader, and can push their own browsers as the defaults on their respective platforms. The most obvious way to make money from a browser - harvesting user data - is largely unavailable to them.
    - BizarroLand 10mo ago
      I would rather firefox release a paid browser with no AI, or at least everything Opt-In, and more user control than to see them stuff unwanted features on users. I used firefox faithfully for a long time, but it's time for someone to take it out back and put it down. Also, I switched to Waterfox about a year ago and I have no complaints. The very worst thing about it is that when it updates its very in your face about it, and that is such a small annoyance that its easily negligible. Throw on an extension like Chrome Mask for those few websites that "require chrome" (as if that is an actual thing), a few privacy extensions, ecosia search, uBlacklist (to permablock certain sites from search results), and Content Farm Terminator to get rid of those mass produced slop sites that weasel their way into search results and you're going to have a much better experience than almost any other setup.
jrjeksjd8d 10mo ago
To be more charitable to TFA, machine translation is a field where there aren't great alternatives and the downside is pretty limited. If something is in another language you don't read it at all. You can translate a bunch of documents and benchmark the result and demonstrate that the model doesn't completely change simple sentences. Another related area is OCR - there are sometimes mistakes, but it's tractable to create a model and verify it's mostly correct. LLMs being applied to everything under the sun feels like we're solving problems that have other solutions, and the answers aren't necessarily correct or accurate. I don't need a dubiously accurate summary of an article in English, I can read and comprehend it just fine. The downside is real and the utility is limited.
- schoen 10mo ago
  There's an older tradition of rule-based machine translation. In these methods, someone really does understand exactly what the program does, in a detailed way; it's designed like other programs, according to someone's explicit understanding. There's still active research in this field; I have a friend who's very deep into it. The trouble is that statistical MT (the things that became neural net MT) started achieving better quality metrics than rule-based MT sometime around 2008 or 2010 (if I remember correctly), and the distance between them has widened since then. Rule-based systems have gotten a little better each year, while statistical systems have gotten a lot better each year, and are also now receiving correspondingly much more investment. The statistical systems are especially good at using context to disambiguate linguistic ambiguities. When a word has multiple meanings, human beings guess which one is relevant from overall context (merging evidence upwards and downwards from multiple layers within the language understanding process!). Statistical MT systems seem to do something somewhat similar. Much as human beings don't even perceive how we knew which meaning was relevant (but we usually guessed the right one without even thinking about it), these systems usually also guess the right one using highly contextual evidence. Linguistic example sentences like "time flies like an arrow" (my linguistics professor suggested "I can't wait for her to take me here") are formally susceptible of many different interpretations, each of which can be considered correct, but when we see or hear such sentences within a larger context, we somehow tend to know which interpretation is most relevant and so most plausible. We might never be able to replicate some of that with consciously-engineered rulesets!
tdeck 10mo ago
Aside: Does anyone actually use summarization features? I've never once been tempted to "summarize" because when I read something I either want to read the entire thing, or look for something specific. Things I want summarized, like academic papers, already have an abstract or a synopsis.
- badbotty 10mo ago
  Haven’t tried them but I can see these features being really useful for screen reader users.
- wkat4242 10mo ago
  Yes. I use it sometimes in Firefox with my local LLM server. Sometimes i come across an article I'm curious about but don't have the time or energy to read. Then I get a TL;DR from it. I know it's not perfect but the alternative is not reading it at all. If it does interest me then I can explore it. I guess I do this once a week or so, not a lot.
  - ruszki 10mo ago
    I highly doubt that no information would be worse than wrong information. Both wars in Ukraine and Gaza show this very clearly.
    - wkat4242 10mo ago
      I just use it for personal information, I'm not involved in any wars :) I don't base any decisions on it, for example if I buy something I don't go by just AI stuff to make a decision. I use the AI to screen reviews, things like that (generally I prefer really deep review and not glossy consumer-focused ones). Then I read the reviews that are suitable to me. And even reading an article about those myself doesn't make me insusceptible to misinformation of course. Most of the misinformation about these wars is spread on purpose by the parties involved themselves. AI hallucination doesn't really cause that, it might exacerbate it a little bit. Information warfare is a huge thing and it has been before AI came on the scene. Ok, as a more specific example, recently I was thinking of buying the new Xreal Air 2. I have the older one but I have 3 specific issues with it. I used AI to find references about these issues being solved. This was the case and AI confirmed that directly with references, but in further digging myself I did find that there was also a new issue introduced with that model involving blurry edges. So in the end I decided not to buy the thing. The AI didn't identify that issue (though to be fair I didn't ask it to look for any). So yeah it's not an allknowing oracle and it makes mistakes, but it can help me shave some time off such investigations. Especially now that search engines like google are so full of clickbait crap and sifting through that shit is tedious. In that case I used OpenWebUI with a local LLM model that speaks to my SearXNG server which in turn uses different search engines as a backend. It tends to work pretty well I have to say, though perplexity does it a little better. But I prefer self-hosting as much as I can (of course the search engine part is out of scope there).
tliltocatl 10mo ago
The thing about translation, even a human translator will sometimes make silly mistakes unless they know the domain really well. So LLM are not any worse. Translation is a problem with no deterministic solution (rule-based translation had always been a bad joke). Properly implemented deterministic search/information retrieval, on the other hand, works extremely well. So well it doesn't really need any replacement - except when you also have some extra dynamics on top like "filtering SEO slop" - and that's not something LLMs can improve at all.
MrAlex94 10mo ago
Looking back with fresh eyes, I definitely think I could’ve presented what I’m trying to say better. On a purely technical play, you’re right that I’m drawing a distinction that may not hold up purely on technical grounds. Maybe the better framing is: I trust constrained, single purpose models with somewhat verifiable outputs (seeing text go in, translated text go out, compare its consistency) more than I trust general purpose models with broad access to my browsing context, regardless of whether they’re both neural networks under the hood. WRT to the “scope”, maybe I have picked up the wrong end of the stick with what Mozilla are planning to do - but they’ve already picked all the low hanging fruit with AI integration with the features you’ve mentioned and the fact they seem to want to dig their heels in further, at least to me, signals that they want deeper integration? Although who knows, the post from the new CEO may also be a litmus test to see what the response to that post elicits, and then go from there.
- yunohn 10mo ago
  I still don’t understand what you mean by “what they do with your data” - because it sounds like exfiltration fear mongering, whereas LLMs are a static series of weights. If you don’t explicitly call your “send_data_to_bad_actor” function with the user’s I/O, nothing can happen.
  - MrAlex94 10mo ago
    I disagree that it’s fear mongering. Have we not had numerous articles on HN about data exfiltration in recent memory? Why would an LLM that is in the drivers seat of a browser (not talking about current feature status in Firefox wrt to sanitised data being interacted with) not have the same pitfalls? Seems as if we’d be 3 for 3 in the “agents rule of 2” in the context of the web and a browser? > [A] An agent can process untrustworthy inputs > [B] An agent can have access to sensitive systems or private data > [C] An agent can change state or communicate externally https://simonwillison.net/2025/Nov/2/new-prompt-injection-papers/ https://simonwillison.net/2025/Nov/2/new-prompt-injection-pa... Even if we weren’t talking about such malicious hypotheticals, hallucinations are a common occurrence as are CLI agents doing things it thinks best, sometimes to the detriment of the data it interacts with. I personally wouldn’t want my history being modified or deleted, same goes with passwords and the like. It is a bit doomerist, I doubt it’ll have such broad permissions but it just doesn’t sit well which I suppose is the spirit of the article and the stance Waterfox takes.

Cheer2171 10mo ago

No, it is disqualifyingly clueless. The author defends one neural network, one bag of effectively-opaque floats that get blended together with WASM to produce non-deterministic outputs which are injected into the DOM (translation), then righteously crusades against other bags of floats (LLMs). From this point of view, uBlock Origin is also effectively un-auditable. Or your point about them maybe imagining AI as non-local proprietary models might be the only thing that makes this make sense. I think even technical people are being suckered by the marketing that "AI" === ChatGPT/Claude/Gemini style cloud-hosted proprietary models connected to chat UIs.

koolala 10mo ago
I'm ok with Translation because it's best solved with AI. I'm not ok with it when Firefox "uses AI to read your open tabs" to do things that don't even need an AI based solution.
- kbelder 10mo ago
  There's levels of this, though, more than two: local, open model local, proprietary model remote, open model (are there these?) remote, proprietary model There is almost no harm in a local, open model. Conversely, a remote, proprietary model should always require opting in with clear disclaimers. It needs to be proportional.
  - koolala 10mo ago
    The harm to me is the implementation is terrible - local or not (assuming no AI based telemetry). If their answer is AI then it pretty much means they won't make a non-AI solution. Today I just got my first stupid AI tab grouping in Firefox that makes zero intuitive sense. I just want grouping not from an AI reading my tabs. It should just be based on where my tabs were opened from. I also tried Waterfox today because of this post and while I'd prefer horizontal grouping atleast their implementation isn't stupid. Language translation is a opaque complex process. Tabs being grouped from other tabs is not good when opaque and unpredictable and does not need AI.
  - enriquto 10mo ago
    What do you mean by "open"? Open weights, or open training data? These are very different things.
    - kbelder 10mo ago
      That is a good point, and I think the takeaway is that there are lots of degrees of freedom here. Open training data would be better, of course, but open weights is still better than completely hidden.

user3939382 10mo ago

Firefox should look like Librewolf first of all, Librewolf shouldn’t have to exist. Mozilla’s privacy stuff is marketing bullshit just like Apple. It shouldn’t be doing ANYTHING that isn’t local only unless it’s explicitly opt in or user UI action oriented. The LLM part is absurd bc the entire overton window is in the wrong place.

PunchyHamster 10mo ago
It's frankly desperate trend chasing from management that lost after starting from near total market domination, and have no idea what to do now.
- takluyver 10mo ago
  > starting from near total market domination That's not really accurate: Firefox peaked somewhere around 30% market share back when IE was dominant, and then Chrome took over the top spot within a few years of launching. FWIW, I think there's just no good move for Mozilla. They're competing against 3 of the biggest companies in the world who can cross-subsidise browser development as a loss-leader, and can push their own browsers as the defaults on their respective platforms. The most obvious way to make money from a browser - harvesting user data - is largely unavailable to them.
  - BizarroLand 10mo ago
    I would rather firefox release a paid browser with no AI, or at least everything Opt-In, and more user control than to see them stuff unwanted features on users. I used firefox faithfully for a long time, but it's time for someone to take it out back and put it down. Also, I switched to Waterfox about a year ago and I have no complaints. The very worst thing about it is that when it updates its very in your face about it, and that is such a small annoyance that its easily negligible. Throw on an extension like Chrome Mask for those few websites that "require chrome" (as if that is an actual thing), a few privacy extensions, ecosia search, uBlacklist (to permablock certain sites from search results), and Content Farm Terminator to get rid of those mass produced slop sites that weasel their way into search results and you're going to have a much better experience than almost any other setup.

jrjeksjd8d 10mo ago

To be more charitable to TFA, machine translation is a field where there aren't great alternatives and the downside is pretty limited. If something is in another language you don't read it at all. You can translate a bunch of documents and benchmark the result and demonstrate that the model doesn't completely change simple sentences. Another related area is OCR - there are sometimes mistakes, but it's tractable to create a model and verify it's mostly correct. LLMs being applied to everything under the sun feels like we're solving problems that have other solutions, and the answers aren't necessarily correct or accurate. I don't need a dubiously accurate summary of an article in English, I can read and comprehend it just fine. The downside is real and the utility is limited.

schoen 10mo ago
There's an older tradition of rule-based machine translation. In these methods, someone really does understand exactly what the program does, in a detailed way; it's designed like other programs, according to someone's explicit understanding. There's still active research in this field; I have a friend who's very deep into it. The trouble is that statistical MT (the things that became neural net MT) started achieving better quality metrics than rule-based MT sometime around 2008 or 2010 (if I remember correctly), and the distance between them has widened since then. Rule-based systems have gotten a little better each year, while statistical systems have gotten a lot better each year, and are also now receiving correspondingly much more investment. The statistical systems are especially good at using context to disambiguate linguistic ambiguities. When a word has multiple meanings, human beings guess which one is relevant from overall context (merging evidence upwards and downwards from multiple layers within the language understanding process!). Statistical MT systems seem to do something somewhat similar. Much as human beings don't even perceive how we knew which meaning was relevant (but we usually guessed the right one without even thinking about it), these systems usually also guess the right one using highly contextual evidence. Linguistic example sentences like "time flies like an arrow" (my linguistics professor suggested "I can't wait for her to take me here") are formally susceptible of many different interpretations, each of which can be considered correct, but when we see or hear such sentences within a larger context, we somehow tend to know which interpretation is most relevant and so most plausible. We might never be able to replicate some of that with consciously-engineered rulesets!

tdeck 10mo ago

Aside: Does anyone actually use summarization features? I've never once been tempted to "summarize" because when I read something I either want to read the entire thing, or look for something specific. Things I want summarized, like academic papers, already have an abstract or a synopsis.

badbotty 10mo ago
Haven’t tried them but I can see these features being really useful for screen reader users.
wkat4242 10mo ago
Yes. I use it sometimes in Firefox with my local LLM server. Sometimes i come across an article I'm curious about but don't have the time or energy to read. Then I get a TL;DR from it. I know it's not perfect but the alternative is not reading it at all. If it does interest me then I can explore it. I guess I do this once a week or so, not a lot.
- ruszki 10mo ago
  I highly doubt that no information would be worse than wrong information. Both wars in Ukraine and Gaza show this very clearly.
  - wkat4242 10mo ago
    I just use it for personal information, I'm not involved in any wars :) I don't base any decisions on it, for example if I buy something I don't go by just AI stuff to make a decision. I use the AI to screen reviews, things like that (generally I prefer really deep review and not glossy consumer-focused ones). Then I read the reviews that are suitable to me. And even reading an article about those myself doesn't make me insusceptible to misinformation of course. Most of the misinformation about these wars is spread on purpose by the parties involved themselves. AI hallucination doesn't really cause that, it might exacerbate it a little bit. Information warfare is a huge thing and it has been before AI came on the scene. Ok, as a more specific example, recently I was thinking of buying the new Xreal Air 2. I have the older one but I have 3 specific issues with it. I used AI to find references about these issues being solved. This was the case and AI confirmed that directly with references, but in further digging myself I did find that there was also a new issue introduced with that model involving blurry edges. So in the end I decided not to buy the thing. The AI didn't identify that issue (though to be fair I didn't ask it to look for any). So yeah it's not an allknowing oracle and it makes mistakes, but it can help me shave some time off such investigations. Especially now that search engines like google are so full of clickbait crap and sifting through that shit is tedious. In that case I used OpenWebUI with a local LLM model that speaks to my SearXNG server which in turn uses different search engines as a backend. It tends to work pretty well I have to say, though perplexity does it a little better. But I prefer self-hosting as much as I can (of course the search engine part is out of scope there).

tliltocatl 10mo ago

The thing about translation, even a human translator will sometimes make silly mistakes unless they know the domain really well. So LLM are not any worse. Translation is a problem with no deterministic solution (rule-based translation had always been a bad joke). Properly implemented deterministic search/information retrieval, on the other hand, works extremely well. So well it doesn't really need any replacement - except when you also have some extra dynamics on top like "filtering SEO slop" - and that's not something LLMs can improve at all.

MrAlex94 10mo ago

Looking back with fresh eyes, I definitely think I could’ve presented what I’m trying to say better. On a purely technical play, you’re right that I’m drawing a distinction that may not hold up purely on technical grounds. Maybe the better framing is: I trust constrained, single purpose models with somewhat verifiable outputs (seeing text go in, translated text go out, compare its consistency) more than I trust general purpose models with broad access to my browsing context, regardless of whether they’re both neural networks under the hood. WRT to the “scope”, maybe I have picked up the wrong end of the stick with what Mozilla are planning to do - but they’ve already picked all the low hanging fruit with AI integration with the features you’ve mentioned and the fact they seem to want to dig their heels in further, at least to me, signals that they want deeper integration? Although who knows, the post from the new CEO may also be a litmus test to see what the response to that post elicits, and then go from there.

yunohn 10mo ago
I still don’t understand what you mean by “what they do with your data” - because it sounds like exfiltration fear mongering, whereas LLMs are a static series of weights. If you don’t explicitly call your “send_data_to_bad_actor” function with the user’s I/O, nothing can happen.
- MrAlex94 10mo ago
  I disagree that it’s fear mongering. Have we not had numerous articles on HN about data exfiltration in recent memory? Why would an LLM that is in the drivers seat of a browser (not talking about current feature status in Firefox wrt to sanitised data being interacted with) not have the same pitfalls? Seems as if we’d be 3 for 3 in the “agents rule of 2” in the context of the web and a browser? > [A] An agent can process untrustworthy inputs > [B] An agent can have access to sensitive systems or private data > [C] An agent can change state or communicate externally https://simonwillison.net/2025/Nov/2/new-prompt-injection-papers/ https://simonwillison.net/2025/Nov/2/new-prompt-injection-pa... Even if we weren’t talking about such malicious hypotheticals, hallucinations are a common occurrence as are CLI agents doing things it thinks best, sometimes to the detriment of the data it interacts with. I personally wouldn’t want my history being modified or deleted, same goes with passwords and the like. It is a bit doomerist, I doubt it’ll have such broad permissions but it just doesn’t sit well which I suppose is the spirit of the article and the stance Waterfox takes.