But only Apple can, at least for now, scan content or your phone. Google cannot. Yet. Small, bit crucial difference.

Frankly, it's more of a problem for my data to be flagged after Google has it than it would be for it to never leave my device at all.

People have certainly run afoul of the law when the data Google has about them is misused.

>Innocent man, 23, sues Arizona police for $1.5million after being arrested for murder and jailed for six days when Google's GPS tracker wrongly placed him at the scene of the 2018 crime

https://www.dailymail.co.uk/news/article-7897319/Police-arre...

We are only talking about what happens when you try to send the data to the cloud, after all. Not about all the data on the device.

You are so close to understand the issue. When Google, or anyone else, is flagging it and sharing the result with, e.g., law enforcement you can have a problem. Apple will no just simply refuse to upload stuff that would be flagged in the cloud, it is flagging it already before upload on your device. And share the results with authorities. That for now they claim to only do so with stuff about to be uploaded is of minor concern, the tech is there to scan all the pictures on an iPhone now. No reason to believe it will stop with intended uploads, or pictures.

> Apple will no just simply refuse to upload stuff that would be flagged in the cloud, it is flagging it already before upload on your device. And share the results with authorities.

None of this is true. A sibling comment describes the process:

>1. Only if you're uploading files are the files matched. 2. Only if the matches are very close are they considered matches. 3. Only if you have multiple very close matches is Apple able to decrypt the low-res versions of the images themselves. 4. Only if a human reviewer discovers any of the decrypted low-res images to be illegal content is any of your information shared with anyone else.

https://news.ycombinator.com/item?id=28120598

Really? Apple and Google have been able to scan content on your phone as long as they have sold phones. They control the OS, therefore they have access to the content. The ability has always been there. The only difference is that Apple has now openly announced what they are going to do and how. If governments wanted to pressure Apple (or Google) to scan content they could have done that anytime in the last few years. If Apple and Google resisted that before, why is now different?

From a legal perspective, there is a significant chasm between "you are hereby ordered to develop a whole new surveillance feature for devices you build" and "you are hereby ordered to add these new (potentially unrelated to CSAM) signatures to your existing surveillance network".

Under what legal basis are certain corporations immune to government demands while others are not?

Google does reporting on how many National Security Letters requesting information about it's users that it responds to annually.

Why doesn't it just ignore them?

> Google does reporting on how many National Security Letters requesting information about it's users that it responds to annually.

So does Apple[1]. Apple gave data on over 31,000 users/accounts based on National Security Letters in the first half of 2020 alone[1].

> Why doesn't it just ignore them?

In the first half of 2020, Apple provided data to the US government (non-FISA or NSL) about 9,000 times, and responded to requests for data with the data about ~85% of the time, and 92% in cases of "emergencies"[1].

[1] https://www.apple.com/legal/transparency/us.html

I don't understand your question. Google, Apple, Facebook et al comply with their legal obligations.

Apple does not have a legal obligation to install spyware on your phone (at least not in the US, not yet).

I think so too but from a different angle. It sounds like the federal government is strong arming Apple to spy on their users. Apple would rather use on-device hashing to preserve some amount of privacy rather than give unfettered access to the feds to their cloud.

On the other hand, Google can run arbitrary queries over your data (published to their cloud) while Apple uses the same algorithm and hash database for everyone (which probably will get reverse engineered and audited by next BlackHat).

What makes you think that Apple will use the same database for everyone?

You can't realistically reverse engineer a hash to source image.

1. Previous submission (http://www.hackerfactor.com/blog/) claims that PhotoDNA is reversible to 26x26 grayscale

2. What’s more important is you can reverse engineer algorithm that’s used locally to compare these hashes. Unlike with cloud-based searches other providers run, this puts significant limitations to what Apple or LEA can do on your phone. E.g. they can’t extract strings from PDF or grep through image metadata or request all images of certain users without knowing what these images would depict in advance. The frequency of hash database updates even if opaque will still provide more insight into what is being searched for compared to current status quo where public has zero insights because all search / aggregation queries are run on the server.

My understanding is they only scan for content if it's being uploaded to iCloud. Scanning on the phone, yes, but only if the image is to be sent over the network.

And they seem to be doing that because they both 1) want to enable e2e encryption for iCloud images, soonish, and 2) don't want to become the world's most popular child porn storage and sharing platform.

I get the arguments for why the fix for #2 is too dangerous to be worth it, but also get why they'd be inclined to choose it.

Apple can already find photos of my cat, of ice cream, and of the sky directly on my phone using AI

A webpage does authentication by downloading a list of usernames and passwords and comparing your inputs in JavaScript in your browser. Another sends a hash out to an endpoint in the cloud.

To the user - using your line of reasoning - they're both the same, right?

This isn't supposed to be a technical comparison, I just want to point out that the implementation matters as much (or more) than the customers' experience.

I don't use the cloud. I don't even sign in an email address on my android phone. That makes all the difference to me.

>Both are equally invasive and both can only be avoided by refusing to use the cloud.

Technically Apple can do a scan regardless of wether or not you posted it to iCloud if they so desired to, or were ordered to.

Technically, Google can sell the entire trove of data they have collected about you to anyone willing to write them a check.

Do you think it's likely they will do so?

The difference is you don't put your data in Google's cloud, they can't sell it. With this Apple change on your local device, you don't have any recourse other than get a different phone.

Also, Google sells your data already don't they? Probably not a great counterpoint.

https://www.tampabay.com/news/2021/05/07/google-selling-user...

Again, the only thing different here is that when you upload a photo to the cloud, Apple scans it on device before the file is uploaded, while Google scans it on their server after it is uploaded.

Nothing is happening to files you don't attempt to upload to the cloud.

Also, if Google is selling everyone's data to third party data brokers, it's news to me.

>Nothing is happening to files you don't attempt to upload to the cloud.

I am aware, but because the scanner now exists on the device, they have the capability of scanning images on your device regardless of whether or not you upload it to the cloud if they so choose to OR ARE ORDERED TO. Do you not see the difference in capability?

“Google has the capability to upload images from your device and could choose to upload every image IF ORDERED TO.”

By that logic, Google is every bit as vulnerable to being ordered to scan all files on Android devices.

So you are saying Google can access your device at any time and pull all data from it, even if you didn't upload it to their servers!? No thanks on either. I'm not sure "other companies can do it too," is a great counterpoint either.

This is classic what-aboutism, yes they are both equally bad scenarios. Apple is just one-step closer with capability already deployed.

There are many photos one could take without uploading or saving them in the cloud. The question is can I safely store things locally on my device without a multi-national corporation being able to set the policy of what's allowed there. The article cites examples of banned memes/images by the Chinese government.

There is also the risk of algorithm mistakes. Without the trust + safety team or any cloud verification, does this just get immediately forward to the FBI?

What is the process to be made non-guilty if the Apple algorithm thinks you have something illegal and you don't if there is no human in the loop involved before it's forwarded to authorities?

At least if you upload to a service someone on the trust + safety team can verify the algo worked properly. I would be very nervous building this kind of system of the risk of false positives! Maybe even so nervous to introduce many more false negatives - which is in itself terrible in the case of something like CSAM.

Here's how it works, in short:

1. Images which are going to be uploaded to iCloud are matched against known hashes. A match has to be very close in order to actually be flagged as a match. The results are uploaded to iCloud Photos along with your image. 2. If your account hits a sufficient threshold of matches (i.e. you have to have multiple/many images matched) then Apple is alerted and they are able to decrypt low-resolution versions of the relevant stored images. 3. Apple's team performs a manual review of the images to ensure that they are in fact inappropriate images and not false positives. 4. If any of the matched images are actually illegal, then Apple reports you to the authorities.

In other words, there are several safeguards in this process:

1. Only if you're uploading files are the files matched. 2. Only if the matches are very close are they considered matches. 3. Only if you have multiple very close matches is Apple able to decrypt the low-res versions of the images themselves. 4. Only if a human reviewer discovers any of the decrypted low-res images to be illegal content is any of your information shared with anyone else.

In my opinion, this is a vastly better system than literally any other cloud provider at the moment; it allows Apple to find and report even more CSAM with effectively no risk of reporting false-positives to law enforcement, while also preserving your privacy through encryption in transit and at rest.

> The question is can I safely store things locally on my device without a multi-national corporation being able to set the policy of what's allowed there.

Didn't Microsoft just start removing people's torrent clients from their computer without permission and announce that you wouldn't be allowed to use your own PC running Windows 11 Home without using an online Microsoft account to log in?