Could someone summarize the problem with Firefox's AI features?

At least when I last checked (months ago), none of those features that involve communicating with external servers would work unless you configure them to (i.e. provide credentials to an LLM provider).

Was I wrong? Have things changed?

▲

baobun 18 hours ago | parent | next [-]

What was your methodology in checking? I got different results using a local mitmproxy on a clean install.

https://sizeof.cat/post/web-browser-telemetry-2025-edition/

	▲	BeetleB 17 hours ago \| parent [-]
		Thanks for the link - I see it's not that much more than Waterfox. Getting to the discussion at hand, which of those pings are AI related? I didn't say FF isn't making network calls.

▲

ksherlock 20 hours ago | parent | prev [-]

I don't use firefox so I can't confirm, but one issue might be 15+ (?!) different config settings needed to disable AI and it still won't go away.

https://news.ycombinator.com/item?id=46095873

▲

BeetleB 20 hours ago | parent [-]

That's a UX issue, but I keep hearing complaints about privacy.

▲

phyzome 17 hours ago | parent [-]

Anything that makes it easy to accidentally send local data elsewhere is a privacy issue.

▲

BeetleB 17 hours ago | parent [-]

> Anything that makes it easy to accidentally send local data elsewhere is a privacy issue.

How is it "easy" if nothing is sent unless you configure the AI?

What I'm asking is: If I do a brand new profile, default configuration, how can any AI related feature send anything that is of privacy concern? If you don't set up an LLM provider, it has nowhere to send to.

I may be wrong, which is why I'm asking in the thread. So far, no one has shown what the problem is.

▲

phyzome 15 hours ago | parent [-]

I have no idea whether any of the AI features require explicit setup vs. automatically use a paid-for API somewhere.

But it also doesn't matter, because that's the kind of distinction that I've seen go back and forth elsewhere.

▲

BeetleB 13 hours ago | parent [-]

OK, to be frank, it seems like people are needlessly crazy paranoid.

I agree with:

https://news.ycombinator.com/item?id=46316763

278 comments, many very angry, and no one can clearly articulate how privacy is being compromised because of the AI features.

On a project whose source is available.

Insane.

	▲	phyzome 4 hours ago \| parent [-]
		Look, it's pretty simple: A bunch of companies have been shoving AI into their products without a lot of consideration for what their users actually want and need. This is a signal for current and future actions that are user-hostile, both directly involving AI and in general (indicative of their current mentality). "Open source" doesn't help when it's a huge project and users aren't actively auditing all of the changes. (And in general we want to trust the developers; if you're having to audit all features, trust has already been lost.)