Restricting AI agents in mobile operating systems platforms from capturing data
signal.orgBy the way here are three wishes for a magic lamp:
1. People stop acting like it matters whether privacy and anonymity are different things.
2. People stop acting like it matters if "Wayland is a protocol".
3. People stop acting like using all your RAM ASAP is not wasting it.
I think it would be unlikely that any spam is going to come from the paid accounts. When phone numbers are way cheaper.
Frankly, the paid accounts are overpriced. But, anyone that wants to pay with money instead of phone numbers is going to be fine supporting Signal anyway.
The best argument for Monero is that the overlap of people that want to pay with a KYC payment method and that would pay in the first place instead of a phone number is very small.
Signal is for pretend-security.
No. Signal is for content privacy by code and metadata privacy by policy and also code but weaker.
If you didn't notice, SimpleX had way less users and way worse UX.
Being built for different things doesn't make it pretend security. You're completely correct that it's not for anonymity. You're completely wrong that it's not secure.
It's also their policy to refuse to update their privacy policy which is currently lying to users about the sensitive data they collect and store.
There are better ways to secure your content and meta data from more trustworthy companies.
Can we please stop it with this?
In real-world practice, the 501c3 named Signal and Amazon Web Services that its servers run on can see the metadata of which users messaged which other users at what time.
This is why they haven't given the users the option to invite other users with a string (or "token" or whatever) so the inviting user can be messaged without a PSTN number or Google id being demanded from the invitee.
If you're trying to buy some good shit on the street or meet up at "da show" with urbanite degenerates, Signal is great.
Serious people, infosec pros, execs, etc stopped using Signal about two years ago -- at least that's what I saw from my corner of the world.
it comes across like more of a ploy to get you to scare people away and get them to use other, less secure, messengers
"Signal" in the previous comment referred to the service hosted by the 501c3 named Signal, not the body of open source software named Signal ; these are overlapping but different things.
FOSS quietly shines in the private (as in, among private entities).
Organizations that are universally recognizable as working on "serious" things have coalesced around Matrix and Element X or other Matrix clients (some not FOSS), instead of services hosted by third parties (such as Slack). Different audience but an example of modern literacy.
"Small" users such as "l33t hax0rs" which I read as "age range of teenager to college freshman", often set up their own XMPP servers and use OMEMO. Perhaps these users are not organized around something "serious", but their choice to have self-respect and gain literacy to act it out is something that can be taken seriously.
To answer your question more directly: anything with decent technical merit that either doesn't require a 3rd party or the 3rd party can be specifically chosen.
I recently wrote on this topic, not so much the technical pieces but the vibe-perspective
https://news.ycombinator.com/item?id=50417774Most likely a decentralized service such as a fork of Signal that uses their own infrastructure, or (self-hosted) Matrix, or XMPP.
Both for account creation, as you mention, and for donations. I see they do take some crypto for donations now, which is good. But not the most obvious one!
Take my money guys!
Good to see they’re making some progress towards addressing this issue and that they are “cautiously optimistic that some major OS providers, like Google and Apple, are engaging with us on how these systems should better interface with privacy-sensitive apps like Signal.”
Rust / WebAssembly can do a lot.
OCR is automatic, high quality, and dirt cheap nowadays, even on device.
[0] https://en.wikipedia.org/wiki/Hardware_overlay#Screen_shots
It is far simpler to try to convince the OS vendor to work with you than to pretend you can fight them with technical means while in the sandbox they gave you.
Screenshots, for example, can be prevented only if the OS provides an API for apps to use to prevent screenshots being taken of sensitive information.
Similarly, if an OS gives LLM(s) access to content as rendered on the screens, via accessibility APIs or similar, that will leak any information made readable to the user unless and API is provided to opt out of that access.
Yeah, it is likely this.
A decade ago I worked for a company that provided secure communications apps (email, IMs, etc) for both Android and iOS and (as one of the Android developers) I remember pointing out to higher-ups that it was trivial to write an app on Android that would tie into the accessibility APIs Android provided to bypass all of our app's network and storage security (the accessibility APIs would allow you to automatically read the text out of any standard OS text controls of any installed app while running as a background service to provide alternative means of presenting the data, like text to speech, etc).
But instead of providing accessibility, an attacker app could just take all that data and silently upload it somewhere.
This did mean the attacker would have to physically access the phone one time to install the accessibility logger service (or they'd have to phish a user to install it themselves), but after that initial install they could forever bypass the security on our app with absolutely no way for the user to know data was being exfiled unless they are the type that is meticulously auditing which apps are installed on their phone.
Destroying accessibility is unfair to people who need it, but I recommended we add a feature to our app that would default to notifying the user with a clear warning whenever another running app had accessibility read access.
My recommendation never got followed, at least for the remainder of my time at that company (they got bought up by Blackberry about a year later and the office I was working in got shut down).
I both don't understand why LLM providers want agentic use via screenshots or screen scraping and am grateful that they didn't decide to just use browser accessibility trees for browser use features. I don't expect we'll be that lucky on mobile where a11y features don't require industry agreement and standard implementations.
Thankfully, Signal pushed and Apple responded pretty quickly. Now we can tighten up our settings via:
Settings > Siri > Excluded Apps Settings > Siri > App Access Permissions
With the option to "always ask," or "always allow," or "never allow."If the default for Lockdown mode users was to send information to the cloud I'd say that was a not just a bad default but a critical issue.
If the general public's first response to "lets put this on my calendar" is "You need to go into 'Settings -> Siri AI -> Security and Privacy' to approve my ability to read from Messages", then that is a huge fail.
If the goal is to ensure security and privacy of a chat screen or an excel window full of bank balances, wouldn't you want to prevent any app, on any OS from getting screen access?
Serious question.
I mean is there a way to prevent all background processes from taking screen shots while, say, Signal or something else private is in use? Across linux, Windows, MacOS, Android, ios etc? I feel like users should be able to set every OS to operate that way.
BTW, I get and understand the failure state that occurs when the person you're talking to is a moron and is running Claude YOLO or something while you're chatting with them. I'm only talking about ensuring security on your own end.
The most secure computing system is an inert rock. It's also not very useful.
> The most secure computing system is an inert rock. It's also not very useful.
Hang on. That might still have user fingerprints. Better lock it up > Across linux
Yes, but you do need to go to great lengths to do this. The reason is because there's so many things that subtly can extract data. If you want to prevent a system that can take screenshots, then you have to disable screenshots entirely. You have to disable screenreaders. That's all possible, but does interfere with user experience, all because you're concerned about someone potentially hijacking that system. Truthfully that's not in most users' threat models. Similarly you'd have to disable all USB ports and capabilities because a USB device can control your system (see rubber ducky). And many more. To give you a small example, go install noscript into your browser (or install the tor or mullvad browser and put on the strict security settings and browse the internet). It increases privacy and security, but it does come at a big UX cost.Meant to credit the EFF there actually, though probably both
AI agents sure are convenient. By looking at your messages they can tell you that there is an accident on the road to the airport your flight is departing from and that you should consider leaving early. So it gets prioritized over privacy.
For people who are against that tradeoff, Apple can make your device less convenient (lockdown mode), I guess that's an advantage over competitors.
> Why the F would any app have access to my signal messages?
How do you think those toast notifications work? That has to integrate with the iOS system, which needs to be able to read (at least part of) the message. While that's default behavior, you can disable, but that requires you not being able to preview your messages. Though different levels of security/privacy for different preferences and threat models. So pick your own.Just give me tech that works the same every time.
Because that is really what you are asking for. Buy a device and an OS for life.
What do you really mean?
The buttons are programmable and nothing ever changes
It’s vulnerable as an old Android OS but I simply don’t do anything important except take phone calls.
That would support standard banking apps?
But a pager-style device is certainly doable if there is a market for it, and probably the best in terms of security. The question is whether or not Signal (the company) accepts it on their network. Normally, they don't allow third party clients.
I had the phone.
It was great.
From my experience, Firefox was pretty bad compared to Chrome at that time, and the mobile app was even worse. It corresponds to the time I switched to Chrome, I only switched back to Firefox later, when they fixed some of their mess and Chrome enshittified.
To me, it was a strategic mistake. Firefox took advantage of the shortcomings of Internet Explorer to become significant, as switching browser is relatively easy. Trying to attack Android is a different story, you can't change your phone OS like you can switch browsers, and Google actually made good software and had a good reputation at the time, and as expected they failed. Not only that but they also seemed to have neglected their browser for too long, so they also lost the only battle they could have won.
If Signal begins making an OS, what will happen to the app? Signal is just a small company, against them are Meta, Google, Apple,... and they don't have the critical mass to hold their users for long. They can't really afford to divert their attention away from their main product. They can however partner with others (someone suggested GrapheneOS) for this.
> you can't change your phone OS like you can switch browsers
There's a chicken and egg problem here. Sometimes you need to target power users first. Think about how the first Tesla didn't go after the everyday driver (like many other EVs). Went after people that wanted fast cars and could afford a high price. They then get established and can bring costs down to make cars that are more reasonably priced.I think Mozilla killed it too soon. They got buyin from the nerds. They should have then used that as evidence that they could ship phones with it preinstalled. Though that's tricky because Mozilla doing anything generates lots of criticism from the community plus if they ship on a subpar phone people will conflate hardware issues with software. Us nerds are hard to please and we kinda just like complaining
That's why you need some way to bootstrap the system. And Mozilla, I think, didn't take it seriously enough. By comparison, Signal did. Signal originally was a pretty good SMS app that provided encryption on the side, so, use Signal for your SMS, and if the other guy has Signal too, then it is encrypted, making the switch easy. I wish they kept SMS support but that's another story. Also one of the few compromises Signal made to privacy is their system that let you know that your contacts use Signal, to create a network effect.
For Firefox OS, I always thought they should have focused on Android first. The idea would be to first make an app framework for Android, something like Electron, but based on the Gecko engine. Now, developers can start developing Firefox OS apps that run on Android, no additional development costs, and judging by how Electron is successful (much to my dismay), it could have worked. This would have served the Firefox mobile app as well.
They probably should have considered at least some compatibility with Android apps on Firefox OS. Firefox OS is Linux, like Android, and Android is open source, they probably could have done it with reasonable effort.
It’s like iCloud. If I send a photo of my kids to my parents, that could wind up on their iClouds. At that point whatever Apple does with that is out of my hands. I can clearly explain this to them if needed. But with AI there’s a lot more opacity, plus the average person does not understand permissions or look at TOS to see how their stuff is used.
Obviously a kids birthday party picture is quite a different risk profile from confidential sourced journalism but always having the thought in mind is a good way to make the sharing a conscious choice to accept the risk or not.
What am I going to do? Not send emails or text anymore to that person?
It’s like glassholes. The problem is when they’re walking around they’re just pushing whatever’s in front of them up to private servers. What am I going to do? Hunt them down and dodge all of their lines of sight?
It may be the best we have, but the bar is so low.
How can you claim to care about security and your desktop application is just a electron wrapper?
Over 300 MB & 1 GB ram for a texting app?
This is a very unfocused team.
And I don't find signal-desktop's load onerous on my middling-RAM computer, though this is the only Electron app I have running (I get the impression that many HN readers have multiple Electron apps going, which I guess would add up.)
- WhatsApp only ever said they used an older version of the Signal protocol (that, e.g. lacks post-quantum encryption).
- I think WhatsApp can encrypt your backups, but that is not the default.
The backup thing, plus the AI stuff mentioned in the article make it much more likely your WhatsApp conversations end up stored in the clear somewhere than your Signal conversations.
The other main reason WhatsApp is currently worse than Signal is that WhatsApp harvests your metadata (to support Meta's ad-tech business) and Signal does not.
But really, the key to privacy is trusting the recipient, because one the data lands on the other device, it's outside of your control, and if the user has malware enabled (consciously or not) there's not much you can do about that.
The smart play is to have companies in many countries and have them federate traffic across each others infra (like tor) and to have third-party clients with their own ability to implement the E2EE bits...
- or, even better: use something like Matrix amongst friends, you only need a slightly technical friend to set it up for you and now it's really hard for someone to violate your trust, and you're not even a big target for mass surveillance: win-win.
As it stands, Signal can still be compelled to act against their users, especially those not under US jurisdiction, and they're forbidden from talking about it. They're directly under the CLOUD act (if they're not a honeypot designed explicitly to drown out other private messengers, which sounds conspiratorial but its the only thing that makes sense based on how effective their marketing is compared to others- the concerns raised against Signal would kill other messengers).
The developers might be bona-fide idealists and the E2EE encryption might be reliable (if only when compiling the app from source oneself), but it feels obvious that the US authorities would gain access to things like contact metadata or forcing a compromised build on request. The app is leaky in this regard, and Signal rents cloud infrastructure which enables covert shenanigans with the server, like compromising the infamously vulnerable Intel SGX which is the only thing keeping people's contact lists (uploaded by default to the server) private.
My general impression is that Signal is about as secure as I can reasonably expect, but that the entire ecosystem is compromised because it's going through two companies that are beholden to the US government.
JWZ site, so if you don't want the HN testicle egg, copy paste, don't click:
It's a stupid take that the country matters. If the tech is sound and trustless, it won't matter what country the company is based in.
Similarly "digital sovereignty" is stupid. E2EE is what matters. And SimpleX shows that you can essentially add highly effective client-side metadata encryption.
For content privacy, you're much better off with an E2EE file service hosted in North Korea than non-E2EE file service in Switzerland.
I think Signal's threat model is more about law enforcement local or national than intelligence agencies with backdoors in the internet.
As it happens a lot of the abuse of our rights comes from law enforcement. So it's a big improvement to protect against that.
I don't think anyone's expecting Signal to be super anonymous when it's run on AWS.
Yeah, I admit they should work a lot more on metadata privacy. Because anonymity is important in messaging for contact privacy.
This is speculation about mass surveillance by intelligence services. If they're in your threat model then Signal is not the right choice.
Single is decent for law enforcement in your threat model, but SimpleX is still preferable.
Technology must work around this fundamental fact, or work within the law: everything else is a perpetually losing battle.
You can't have an organisation providing anonymous E2EE if there are laws that force KYC or CSAM scanning; and you will be hard pressed to argue on the side of fraudsters and kiddie fiddlers in front of the population.
Like I said, the best solution is small local groups; just like neighbourhoods- there's no actual need to have ONE directory and ONE "instance" of a chat network outside of convienience.. and truthfully, it's not that convenient, because everyone chooses a different chat network anyway. (Whatsapp, Signal, Telegram, Facebook messenger-- which isn't whatsapp for some reason, Wechat, Threema, etc,etc,etc,etc).
Ah yes, I do remember: https://soatok.blog/2024/08/14/security-issues-in-matrixs-ol...
As for Signal being open source (or your claim in the link that it is not open source as in they don't provide the source for builds; and you only need to look at the client source for the security claims), as of a previous post about Signal on HN I have started verifying the builds on iOS which I can confirm are fully source reproducible:
https://github.com/commandersaki/signal-ios-build-verificati...
Signal already provide instructions on Android reproducible builds and it wouldn't really take much to extend that to fully source reproducible.
Most people admire German engineering thanks to the "Deutsche Qualität" ad campaign, yet we forget that came from Opel, one of most unreliable carmarkers at the time.
Apple has invested billions to advertise how private they are, but reality is more complex than that. They have been hostile to security researchers multiple times, and have been slower than the industry average at patching vulnerabilities multiple times.
This is unfortunate because a lot of the people who are knowledgeable and care about security/privacy spend their days in front of a keyboard, which is always going to be vastly better to jot out messages with than any kind of handheld mobile device.
It's certainly not. A reasonably hardened Linux box, preferably running open firmware, is certainly more private, and also probably more secure, at least against interference by Apple.
The GrapheneOS store has less than 20 apps and does not include third-party ones, just system apps.
A Graphene user should install from the play store.
So they do support Graphene and do not warn people not to install Signal on Graphene.
If you don't want to use the play store, not even anonymized via the Aurora store, then you have a special case, and this special case apk exists just for you.
It doesn't have to, signal runs on AOSP and doesn't enforce the fake "security" theater Play Integrity scam.
So why should it "officially support" GOS?
> they are not on their app store
Why should Signal be in GrapheneOS app store? Their app store is intentionally so minimal, there's no good reason do add the unrelated, optional app if Signal doesn't restrict installations from Play Store or Aurora.
Signal can work on being added to Accrestent if they want to be promoted, but that's unlikely given that Signal has a ~hard reliance of GCM for notifications (and apparent refusal to adapt Unified Push as an option).
Signal is aiming for being a mainstream, not niches like GrapheneOS, I'm afraid.
Which it does really well, and it is actually the only non-BigTech, E2EE messenger app that remotely achieved that.
> not niches like GrapheneOS
I don't think GrapheneOS wants to be niche? If you install the Google Services on GrapheneOS, normies genuinely cannot see a difference with any other Android they have seen in their life.
The issue is telecoms are some of the most devious monopolists you will ever encounter. They literally helped change laws, add certification barriers, assist 3 letter organization campaigns on US citizens, and self-sign firmware images... blacklisting 3rd party OS on hardware though imei anti-theft systems.
Starlink is successful because they pulled off a miracle business deal getting these telecoms to cooperate with Direct Connect satellite VoLTE services. =3
https://krebsonsecurity.com/2019/08/who-owns-your-wireless-s...
GrapheneOS is not a direction in that, you're just trading one megacorp for different not-yet-megacorp. Both have security-first manifest, which means both insist on making the devices secured from their owners.
They allow you to isolate google-apps - great.
They don't allow root - meh, I guess, but I'm not really trying to do real work on a phone (or I'll just use termux).
The fact you can't change the lowest levels of the operating system might be annoying if you're and Android developer (idk) but as a user its not so bad.
The great thing is they are open source. If a group of like minded people wanted to distrubute GrapheneOS but hackable I'm sure they could if they had the expertise and resources.
And if GrapheneOS somehow becomes malicious I'm sure someone will fork them and form a like minded community around different values.
That's not really true. We don't _support_ rooting and discourage it, but people definitely do root GrapheneOS. There are also guides on how to do it, etc. I've even seen people do some other weird stuff by making changes to the official images and re-signing them.
For many people, they have no device other than their phone on which to do real personal work.
That somehow we've accepted that a super-powerful pocket computer is just a slab for scrolling TikTok and running banking apps is a _tragedy_
Instead of shrugging we should be rebelling. Reject attestation, reject users being treated as hostile actors on their own devices, reject being told you can't be trusted to control your computer.
Strong point. Somewhat weaker, but also relevant, is that foldable phones at this point cross strongly into "devices for doing real work" space (a tablet sized screen with logistics of a regular phone does a lot here), and as a user of one[0], I feel increasingly limited by the bullshit limitations the platform and app vendors introduce to lock the user out of their own device while keeping privileged access for themselves.
So if you don't care about the poor, there's also now the ones with expensive phones to cater - another reason to stop treating mobile devices as TikTok consumption devices.
> Instead of shrugging we should be rebelling. Reject attestation, reject users being treated as hostile actors on their own devices, reject being told you can't be trusted to control your computer.
Very much this.
--
[0] - Galaxy Z Fold7 + small Bluetooth keyboard and a foldable mouse for heavier work, have made my personal laptop redundant; I haven't turned it on in over a year now.
Instead of making a stance against this stupid anti-competitive practice, you decided you want to play along, relying on the goodwill of the community to push app developers to allowlist you.
> Instead of making a stance against this stupid anti-competitive practice, you decided you want to play along
Again, not sure what you're talking about, but if you are talking about Play Integrity, then that couldn't be farther from the truth. We've been talking about this issue for a long time now, calling Google out on it, and even talking with regulators.
> relying on the goodwill of the community to push app developers to allowlist you
This also doesn't make any sense. Apps don't have to allowlist GrapheneOS for them to work.
Banks don't have to use remote attestation, they choose to. Don't blame GrapheneOS for that.
clearly that's false as it needs to be signed by the organization to be of much use as a day-to-day device.
The comment you answered to does not mention building your own GOS?
> That is very inaccurate. GrapheneOS is open source and people are more than welcome to make their own custom builds. Although we don't support rooting, many do. That's up to them.So... I don't know what you are saying. That because you cannot have root AND your bank app, therefore GrapheneOS... what exactly?
Banks don’t accept running on GrapheneOS even without root. Hardware attestation hurts them. They ask their users to complain to their banks, leave reviews on the app store and be loud about it. To end this practice that violates the users’ freedom? No, just to whitelist GrapheneOS, fuck everyone else.
They just want another key to the walled garden.
In practice, it doesn’t matter if it’s open source, it doesn’t matter if you can root your phone. These are just marketing words to appeal to tech-savvy users. Because in practice, trying to exercise your freedom means you have a brick in your hands.
I, too, am pissed at banks that require remote attestation. Or goddamn Netflix & co with their DRMs. But that's not a reason to write nonsensical rants all over the place.
Go back to your Google-signed Android or iOS, I'm sure you'll be happier there.
The point is that GrapheneOS does not give you control. That’s it. No rant. Use your alternative OS all you want, and when it stops being aligned with your needs, you’ll be stuck in the same place you were with Google-signed Android and iOS. How’s that control?
It is very much a different place.
> How’s that control?
Well it's a tradeoff. Your alternative to not watching DRM-ed Netflix is to not watch Netflix, that's all there is.
This argument (or the equivalent "just fork it") gets trotted out whenever anyone has a complaint about the company behind any open source software.
It's bunk. Sure, if we're talking about a library with a couple of thousand lines, just fork/build it.
But we're talking about an entire operating system here. Building it (and then the even harder part of getting it installed on your phone) will be beyond the capabilities or interests of most people, even with agentic help. And then presumably if you're doing that, it's because you want to actually change something non-trivial, in which case even for an interested, reasonably skilled person with agentic help I would say were talking about a week to potentially months of work, not to mention the ongoing upkeep after that.
Companies even want you to think that them securely holding all your data is the same as it being private to you. That's on purpose.
(Fun fact: Signal only recently enabled registering without sharing your fucking phone number just a few months ago. and only on Android. They don't need your phone number, they don't use it for anything. and you wouldn't want them to. But they want it anyway. That's not privacy.)
Apple is one of the only companies that truly cares about privacy over mere security. This doesn't mean they don't also have some truly serious security, but rather that they simply don't collect data they don't need, and also that they avoid collecting anything other than what they need.
Contrast this with other tech giants, who want all your data, but promise to keep it secure.
I have family conversations in Signal that I want Siri and Spotlight to be able to index. I've been considering moving to a different app solely because of this. Make it a setting!!
I don't trust Google and Facebook, but I do trust Apple. Don't force-distrust Apple on my behalf, I don't want that. This should be my choice.
Signal could even make it the default, because I have no shame in asking my family to let me index their chats. That would be better for the masses, I'd imagine.
Historically this has always been the case and it's unclear why something that is essentially impossible should become an expectation in anyway.
[1] - https://en.wikipedia.org/wiki/Messaging_Layer_Security
[2] - https://en.wikipedia.org/wiki/OMEMO
[3] - https://en.wikipedia.org/wiki/Off-the-Record_Messaging_proto...
Over the years we’ve had various requests for conversation or personal usage data, including a court case of some church, and we told them the truth - that we don’t receive any. They said to send that in writing and that was it.
Here was my reply years ago to Moxie Marlinspike about why decentralization is key to provable privacy:
https://community.intercoin.app/t/web3-moxie-signal-telegram...
The problem is that this is also what someone not serious about protecting our privacy would say.
Similar claims from big tech ad-companies I support... taking with a grain of salt.
This decision is mine to make, not Signal's. As the owner of the computer, I decide what programs have access to what. Signal has no business attempting to block my own user agent, which I run with my fully informed consent, accessing my own data.
Preventing man-in-the-middle is different, something you can actually do technically. That is something you can actually promise to a degree.
Either Signal respects that regular people have general purpose computing or they don't.
As a user of signal, I do expect signal to prevent my chats from being read by anything other than the other contact.
Signal is better than basic SMS, iMessage, Facebook Messenger, etc. That's got value. It's convenient and has reached critical mass so the barrier to entry is low enough that it's a viable option.
It's better than nothing, and just secure enough to be worth it and just secure enough to maintain a customer base.
It's still a business, and they will make it as secure as needed to earn money, and no more secure than that.
[0]: https://github.com/signalapp/Signal-Android
The one you download also has proprietary binary blobs.
This is straight up lies, and I don't need to prove it to this specific user, because I already have, but (s)he likes to be obtuse and wilfully ignorant.
For anyone else, full source reproducible verified for iOS: https://github.com/commandersaki/signal-ios-build-verificati...
And the Android clients come with closed source components from Google. It's still not a problem, but it's true.
The build of Signal Server from public source does not mean that's what they deploy in production.
For security analysis the Signal Server is immaterial, all claims of security can be verified from the client. For all intents and purposes the server is just a dumb pipe.
Then why withhold the code?
What can Signal do to annihilate your concerns without giving you access to their system?
At best they (very) passively fight against third party clients connecting to their official servers.
At best they (very very very) passively fight against self-hosting the server by not providing elaborate guides and having the public repo lag behind production.
"Signal's model means you don't have to trust the server" is a meme post, but it hasn't been accurate since the early 2020s.
You can't just say it and leave. Anything better than SMS which was not even meant for communication.
But lumping imessage and even Facebook messenger in the same pile is disingenuous.
There are serious cryptographers working on iMessage too, eg. https://securitycryptographywhatever.com/2024/03/03/post-qua...
SMS should indeed not be included in the same conversation as those two.
You have no idea how demotivating your comment is for people who have taken a pay cut, and dedicated their lives to try and find a better balance between usability, privacy and sustainability.
There are better alternatives to signal and other companies who aren't outright lying to their users about what data they collect and store in their privacy policy.
You can read more about the issue here: https://web.archive.org/web/20250117232443/https://www.vice....
Note that at the bottom of that article they talk about giving users a way to "disable" the pin, but there is no actual opt of the pin. If you choose not to create one, one is created for you, and that's used when uploading your data to the cloud even though you won't ever be able to recover the data. You cannot opt out of the data collection.
I don't think that signal is too incompetent to keep their privacy policy updated. They were doing it regularly before this change. I don't think they're so ignorant that they don't understand why it's extremely important for a service like theirs to be transparent and truthful in their policy either. It's looking like either the company is intentionally misleading people who are in very sensitive situations (including activists and whistleblowers whose lives and/or freedom can be at stake) or they're doing their best to warn people away from using their service as loudly as they can by not updating their privacy policy.
In either case, because I continue to see people using and recommending signal while unaware of the new (now years old) data collection practices and/or signal's failure to inform users of their actual risks while using the app, I do my best to make sure more people are aware so they can make an informed choice for themselves according their circumstances.
If Lorenzo is cool with the data signal collects and keeps in the cloud, and how they're doing it, that's great. Every user should have that opportunity, but right now it takes a fair amount of digging to uncover the lies being told in the polices signal has published on their website.
That said: periodic reminder that Signal cannot protect you from your chat partners' devices. You might have this protection enabled, but if your chat partner has it disabled, their OS's AI can now read your chats.
if it was, it wouldn't require a phone number or a payment to sign up. that's an utter non-starter for a privacy tool. no public API, no third-party clients, no first-party web client. Signal Messenger LLC is, painfully obviously, a SaaS company operating at a loss and waiting for a buyout from a megacorp.
and the undeserved glazing it got from the media reeks of operation mockingbird bullshit.
The payments use zero-knowledge proofs, so they're not linked to you. All that's known by anyone is that your payment method purchased a Signal account.
There are third party clients.
So you're advocating for web cryptography. That opens up a significant attack vector for your privacy.
It's a non-profit that can't go against privacy for its users.
>and the undeserved glazing it got from the media reeks of operation mockingbird bullshit.
Instead of spreading fud, since signal clients are fully reproducible, meaning the code on github is what's in the apps running on your phone, please point to the code that makes it backdoor in some way. Because you could easily prove these claims. But instead you provide no evidence.
it links your account to a government issued ID in most countries.
>All that's known by anyone is that your payment method purchased a Signal account.
We kill people based on metadata.
>There are third party clients.
there is no public API. any platform can have the kind of frankensteined clients that reverse engineer the official one. the point is that if they cared about privacy like the OP claims, the official client would be a suggestion.
>So you're advocating for web cryptography. That opens up a significant attack vector for your privacy.
are you under the impression that web cryptography uses random npm libraries for cryptographic primitives?
>Instead of spreading fud, since signal clients are fully reproducible, meaning the code on github is what's in the apps running on your phone
is the server code on github also?
https://github.com/signalapp/Signal-Server
How would you be linked to your phone number if you used usernames and changed them after connecting with the contact as they suggest?
Web crypto: https://www.devever.net/~hl/webcrypto
Just as we don't know exactly what's running on a signal server, that's the whole thing with web crypto.
Maybe a public API matters for user freedom, but it doesn't matter for user privacy.
>We kill people based on metadata
Again, signal is not a highly metadata private messenger. So this is a real concern, I agree. But it's for a threat model with more advanced adversaries than national law enforcement like the FBI in the US.
They do take significant steps to minimize metadata. But it's not anonymous.
and that's exactly my quarrel with it.
it could've been significantly more anonymous if it was a privacy tool. but it is an "app", and the people behind it want to make money, not to champion privacy.
About two of your points:
- Either a phone number, or some kind of payment is the best way to avoid being overrun by bots and scammers making huge volumes of accounts. We're living in an age where people can't even run a web page about knitting without being scraped to death, and that's not even by scammers. You want Signal to let anyone make an account, which can burn resources with audio and video calls, without incurring some kind of nontrivial cost to themselves?
- It's been well-explained why they discourage third party clients (it will interfere with fast development of the service). That's disappointing, but OK for now? What they are doing seems to be working to get a lot of ordinary people to use it.
FYI a while back (years ago), I built Signal Desktop from their source, and had no trouble connecting and using that as my daily driver Signal client. So at the very least, they tolerate that.
"Settings > Notifications > Notification Content and selecting "No Name or Content" (or "No Name or Message" on Android), the device prevents the operating system from storing sensitive message previews in its internal database. Additionally, enabling Sealed Sender in Signal's privacy settings hides the sender's identity from Signal's servers, preventing traffic analysis."
Until / unless the recipient responds and the traffic analysis detects back-and-forth and pairs your message's recipient with the recipient's recipient.
https://abcnews.com/blogs/headlines/2014/05/ex-nsa-chief-we-...
If you need to reduce spam sign-ups, let me pay a few dollars in crypto instead.
Now a Signal account will be linked directly to your Google/Apple account and credit card.
I for one 100% agree with you, they should accept crypto, regardless.
https://signal.org/blog/private-contact-discovery/
https://signal.org/blog/signal-profiles-beta/
https://github.com/signalapp/ContactDiscoveryService-Icelake
https://signal.org/blog/private-contact-discovery/
https://signal.org/blog/signal-profiles-beta/
https://github.com/signalapp/ContactDiscoveryService-Icelake
The new "we collect your contacts and profile data and keep them forever in the cloud" feature is not the same thing. It is not optional. There is no way to opt out of, or prevent the data collection. It doesn't hash and collect all the contacts from your phone's address book like contact discovery does, instead it stores a record of the people you communicate with while using signal. I'm sure can see why that would be extremely sensitive information for groups like activists, journalists, and whistleblowers.
It also uploads and permanently stores your profile information in the cloud. This is very different from how the profiles feature was introduced. From your link: The server doesn’t have access to any profile content because it is end-to-end encrypted and shared via the same Signal Protocol messaging channel that already protects your conversations and calls. Now, instead of just transmitting that data over the messaging channel, that profile info is uploaded, stored, and sitting on cloud servers. If you are a signal user and you weren't even aware that all of this was happening, that should tell you all you need to know about how trustworthy signal is, but it gets worse.
All data collected and stored is also kept using (a notoriously leaky) enclave and it's protected with a pin. (see https://web.archive.org/web/20230519120156/https://community...)
Their excuse for this new data collection is that it exists for backup purposes. However, they could have offered a local non-cloud backup option or simply made the backup feature entirely optional (and signal users asked for that, in many threads and comments around the time this rolled out, see for example https://community.signalusers.org/t/dont-want-pin-dont-want-...) but signal refused to do that.
In fact, in the face of user pushback against the mandatory pin nonsense and growing concern from users and security experts about their new data collection scheme (see https://web.archive.org/web/20250117232443/https://www.vice....) what signal did was to allow users to opt out of setting a pin, but all that did was auto-assign a pin in the background which was then used to upload all the same data to the cloud, even though because the user then has no idea what the pin was auto-set to, they could never use it to recover their profile settings or signal contacts. One of the few things signal was clear about was that by opting out of setting a pin you could never recover your profile and contacts. Uploading and storing the data anyway makes it pretty clear that the data collection was the goal all along.
As far as other actors who do more to protect users in this space there are many options. The one I've turned to for secure messaging is Jami, and while it admittedly lacks a certain amount of polish, I've had no trouble getting non-technical people set up with it for secure messaging, file transfers, and video calls.
Of course Signal's biggest sin is lying to users about what their risks are when using the app in their privacy policy which was never updated to include any information after they started collecting and storing sensitive user data in the cloud. Not even when they later added yet another feature (optional this time) which stored the whole contents of messages in the cloud. It's my suspicion that refusing to update their privacy policy is their way of warning people away from the service as loudly as they can, but either way signal is clearly untrustworthy.
what a sad day for hackernews
Signal feels like a begrudging partner in this, not someone who has your back. I love Signal but they have a lot of work to do. It was only recently that (and too late for me) that they could even backup the last 90s of chats when sending in a device for repair. It feels like they hate users. Why is WhatsApp (which harvests your entire address book to meta) still have a better experience?
Another similar thing recently was I was trying to help a woman send a document to her lawyer, and for whatever reason I couldn't figure out how to share it with him via text or whatsapp. Somehow I was able to share it to my own phone via whatsapp, and from there I made a group text in Whatsapp with me and her lawyer to share the document.
With AI agents I think we're similarly getting into territory where it's not so clear-cut as Signal is making things out to be. I share the knee-jerk "keep your agents out of my stuff!" reaction but I don't think that's a sustainable/pragmatic choice. It might be correct as a principle but if it locks you out of the operating system, it's not Signal's job to disable key OS features on the grounds that the OS is insecure - at a certain point Signal might as well just wipe the hard drive, but that's not Signal's role. Signal is there to enable people to send messages, not stop them from doing so.
I find this surprising. What mechanism enforces this? This kind of restriction, AFAIK, is only possible on Android and iOS because users do not control the system (no "root" on user builds). On Linux, I certainly can screenshot Signal or any other "app" because the whole OS is under my control.
ps. Yes, trust is an onion, blah blah HDMI decoder boxes, etc, but I’m limiting the problem space to ‘software solutions only’, matching the problem space that the vast majority of AI device scanning of Signal content occurs in.
I would stop using Signal if it required "attestation," which my GrapheneOS phone does not have anyway.
https://grapheneos.org/articles/attestation-compatibility-gu...
It's not even hard to imagine why this is useful. Imagine you're sharing/streaming your screen and your password manager accidentally opens on the shared area.
With the biggest one being the number of people who apparently assumed that when they send a message to someone, that person couldn't just tell anyone else what they said anyway.
Maybe it's just because it's less widespread where I am, but nevertheless.
I hope they will add anonymous crypto support, but you can create an anonymous signal account.
you can never be confident that it won't be taken from you when the number is inevitably reused.
there was once a way to prevent that from happening but now you won't even get a notification that it's happening so you won't have a chance to move your number.
it's less hassle to just never use signal.
In the countries like Russia they simply blocked incoming registrations SMS from Signal and the others to wade off people from the service. I wouldn't risk using a direct payment in a country famous for jailing people for payments to "extremists" organisations either.
Just do email, or stop calling yourself "private" and "secure"
I do think they should accept crypto.
Most likely such an attempt will fail even earlier than Google/KYC become relevant: their account creation vCaptcha de-facto doesn't accept challenges made from VPN connections (based on app reporting the network status, not endpoint IP).
Of course you can?
Maybe I am missin your point, but that is fundamental basic that you put the encryption and decryption in the ends, and you can transfer over any inherently insecure technologies.
But you need the ends to be secure of course.
I should be able to expect basic privacy in my day-to-day chats without needing to spend time/money/resources setting up a goddamn second device just to have truly private chats with my friends and family.
Cry me a river.
[1] or instead of your phone number, give us your CC number.
I want them to accept anonymous payments too, but they use zero knowledge proofs, so the non-anonymous payment does not link you to them. Also, they don't get cc number, Google does.
Whether this fact is a pro, con or irrelevant to protecting software user "privacy" is a question left to the reader
Different readers may arrive at different answers. Not all readers will necessarily share their answers in comments. Commenters comprise a small portion of readers
There's no need to decrypt anything when other parties can compel the corporation to remotely install and run executable code on a particular user's computer at any time ("updates", RCE)
https://raw.githubusercontent.com/signalapp/Signal-Android/m...
"Finally, it's time for the moment of truth! Let's compare the APKs that were pulled from your device with the APKs that were compiled from the Signal source code. The [`apkdiff.py`](./apkdiff/apkdiff.py) utility that is provided in the Signal repo makes this step easy.
Using this script to check the APKs is helpful because APKs are compressed archives that can't easily be compared with a tool like `diff`. The script also knows how to skip files that are unrelated to any of the app's code or functionality (like signing information, or extra harmless metadata added by the Play Store)."
Signal wants people to verify their Android APKs with a script provided by Signal (cf. https://raw.githubusercontent.com/daniellockyer/apkdiff/mast...) "that makes this step easy"
Perhaps the reason someone might be trying to verify the APKs match is that they don't trust Signal
2018
https://apps.irs.gov/pub/epostcard/cor/824506840_201812_990_...
2019-2024
https://projects.propublica.org/nonprofits/organizations/824...
2024
Signal Technology Foundation
Compensation
Key Employees and Officers Compensation Related Other
Meredith Whittaker (Director / Pres Of Signal Messenger) $743,531 $0 $51,697
Surfrider
Compensation
Key Employees and Officers Compensation Related Other
Chad Nelsen (CEO) $251,790 $0 $17,872I cannot expect privacy on Signal for voices and opinions critical of Israel. I wouldn't trust any marketing about privacy coming out of Big Tech, YCombinator or Silicon Valley. They made their bed and now they have to sleep in it.
* Discuss the issue with google keyboard privacy
* Discuss the issue with the google translate API
* Release all of the backend automation code for their infra that runs signal
When a company lies to their users about their risks when using an app being marketed to activists and whistleblowers whose lives/freedom could be put at risk you know exactly how they feel about protecting their customers.
I don't see the same privacy policy online.
> Signal Terms & Privacy Policy
> Signal is designed to never collect or store any sensitive information. Signal messages and calls cannot be accessed by us or other third parties because they are always end-to-end encrypted, private, and secure.
It's _designed_ to never collect...
> Signal utilizes state-of-the-art security and end-to-end encryption to provide private messaging and Internet calling services to users worldwide (“Services”).
That is the first sentence I see under "Privacy Policy".
> Your calls and messages are always encrypted, so they can never be shared or viewed by anyone but yourself and the intended recipients.
Is this the part that is a lie?
> Information we may share
> Other instances where Signal may need to share your data
* To meet any applicable law, regulation, legal process or enforceable governmental request.
* To enforce applicable Terms, including investigation of potential violations.
* To detect, prevent, or otherwise address fraud, security, or technical issues.
* To protect against harm to the rights, property, or safety of Signal, our users, or the public as required or permitted by law.
Oh, I see what you mean. That's pretty broad.
That changed a long time ago when they started collecting and storing sensitive data in the cloud. Exactly the kinds of data they were getting legal requests for. (see https://web.archive.org/web/20250117232443/https://www.vice....). You can fake your name, use a dedicated phone/number, and never upload a photo but your social graph is always highly sensitive data.
Worse than that they are storing that data in not very secure ways (https://web.archive.org/web/20230519120156/https://community...) and not being very clear at all with users about any of this. At the end of the vice article they talked about adding an option to "disable" the pin, but what they actually did was give users a way to opt out of setting a pin, but that just auto-created a pin for them and then still uploaded all the data to the cloud. There is no way to opt out of them collecting and storing your data.
If the lie at the top of the page titled "Signal Terms & Privacy Policy" isn't bad enough, their actual privacy policy has also not been updated since before they started with the data collection/cloud storage, so none of that is included in the "Information you provide" section of their privacy policy. That just talks about the optional "contact discovery" feature. They later added another optional feature where you can store actual message content in the cloud too, but still never updated their policy. They say "Signal queues end-to-end encrypted messages on its servers for delivery", but at least in some cases that data is stored in the cloud (also not their servers) forever as a "backup". This also means the "We will update this privacy policy as needed so that it is current" line in their policy is also a lie.
My best guess is that the company refuses to update the privacy policy as a dead canary to warn others that the service is compromised as loudly as they can (in the US gag orders can stop you from saying things, but can't force you to say something - or in this case to publish updates to your privacy policy). The alternative explanation is that they're just intentionally misleading people who have a lot on the line about their risks when using the service.
In addition the new owners would probably undermine the encryption to curry favor with the government.
No Signal for me. I'll stick with WhatsApp.
There are still a lot of snafus with WhatsApp like it not using an end-to-end encrypted back-up by default. They make you jump through all sorts of hoops to enable it. I find this fishy since there's always someone in your conversations who hasn't enabled E2E back-ups, enabling LEA and others to read it (eventually). And for some strange reason they're requiring a separate password for the back-up.
All of your argument is pure conjecture and untrue. But the last line is the gold. You'll actively choose and defend using a messenger from the most willing company on the planet to sell you out as your validation. Ignorant thought process said out loud being normalized in 2026 is wild.
In addition billions (including 300 million Americans) are using it and the U.S. government would never allow any other nation's intelligence or LEA organization to hack it. I therefore consider it as safe or safer than most alternatives.
How do you imagine the Signal Foundation would run the service if people started using it as actively as WhatsApp? And with the numbers of users WhatsApp is currently serving? They simply don't have the money or expertise to do that. They'd have no choice but to sell it or curtail the service.
And WhatsApp has the added advantage of having a very large and politically connected corporation behind it.
"Now Moxie you put in that backdoor or we'll put you in a hole so deep no one will ever find you." If they'd threaten Zuck like that all Hell would break loose in the White House.
- demands your phone number or a payment method
- actively hostile to independent reproducible build and distribution efforts by projects like f-droid
- publicly stated they prefer proprietary app store installs for user metrics
- built metadata privacy on Intel SGX knowing data is easily exfiltrated
- Federation is forbidden forcing all global metadata rich network traffic to one infra stack.
- Almost every Signal binary on the network is signed by one key, with no publicly accountable HSM to prove it is not signing privacy compromised binaries on the side served to specific people
Centralized authority will never lead to good privacy outcomes.
I use Signal but it annoys me that I can't use some kind of permission based system to give my agents access to my chats.
That's a pretty messed up take unless all of the people you're chatting with gave their explicit consent.
> My agent
Are you Sam Altman or Dario Amodei?
Why?
Historically it has always been accurate. If you send a letter to someone then you no longer have control of it. Same with email.
It's not like this behavior is excusable outside of Signal, but couldn't you take a hint when someone is communicating with you over a privacy-focused messaging app?
Also, I am mostly talking about privacy, not systems security.
If you build a messaging app to protect your users against the NSA, then it won't be accessible to most people. If you build a messaging app that has all the features all the users want and "just works", then it won't be very private (see Telegram).
The whole game is to choose what you solve. Signal has been extremely successful at that: it is trading some amount of convenience for more privacy, and some amount of privacy for more convenience. Signal literally put E2EE on the map for normies. Saying that they are not serious about privacy just shows how much you ignore about how security works. It's not just you: see all the people who couldn't explain how the Signal protocol works (even at a high level) but are the first ones to say "I cannot tolerate it because it requires my mobile phone, and my personal threat model is that I am obviously so important that leaking the fact that I use Signal would put me at risk".
"I won't use it unless they remove the need for the phone number" or "it is not open if they don't enable root access by default" is the equivalent of talking for hours about the unboxing experience of a product because that's easy to do: it's done by the people who actually don't have anything interesting to say about the product because they just don't understand how it works or how it's built. So they talk about the colour of the box. And they have very strong opinions about it.
But I really hope Signal knows it comes across as disingenuous and hypocritical to insist they're serious about protecting user privacy while routing every push notification through a service offered by a documented NSA partner (Google Firebase). This is a huge metadata leak to the adversary that motivated TextSecure to be created in the first place.
I appreciate the accessibility to "normie" users, but give power users the option to set up their own UnifiedPush distributor, the way Molly does.