Even if they encrypt it, the computer needs access to the data thus needs the decryption key. So it's not very secure anyway.

Even if it were encrypted, if access to it doesn't involve explicit confirmation and a password then it can be automated.

And if it can be automated then malware that gets on the machine will be able to access it whether it's encrypted or not.

But let's be real, the whole reason Microsoft is doing this is so they can parse your data for AI. And storing it unencrypted makes it easier for them.

Also "the data won't leave your machine" is a red herring. Yeah the data won't; but the results of AI processing will. They'll take what they need and transfer that out, and leave you holding the bag.

Nmap is a "hacker tool" and all it does is ask computers what ports they have open, something they are set to advertise to the world.

This is a "hacker tool" in the sense that it is accessing data in an unintended way, in the same contect as nmap using protocols intended to communicate for a set purpose to built a list of possible attack vectors.

I wouldn’t really call it a hacker tool any more than you would call a hammer a thieves tool.

IANAL, but I’m pretty sure a hammer is a thieves tool if used in the commission of a burglary.

Those devices used by employees to remove security locks from CDs/DVDs aren’t “thieves tools” when used as intended, but when my dumb ass got caught with one while stealing from Blockbuster, the judge considered it one.

You do have a point, but it does highlight why Microsoft's framing is bad.

Microsoft is basing their approach to this on the concept that your MS account-secured local machine is itself secure, so whatever is in it is fine, because hey, your confidential work info is probably also in your hard drive and unencrypted, so if a bad actor can steal the pictures of it, then it can also steal the original document.

Which mostly is true, to be clear, but it fundamentally misunderstands how much juicier and easier of a target is a reliable, searchable database that logs all activity stored in a consistent location, as opposed to potentially having to extract everything up front. Plus, even if there are few guardrails to all data inside your system, there are some, as this will likely include info you may keep hidden, password-protected or encrypted both locally and remotely. There's a reason my password manager asks for my credentials manually once every time I use it.

With easily searchable text, search for "bank" and get all accounts login. Yay no need to wait for the hacked user to get on his banking site he's been there before. Quick in and out without being noticed and you got all you need to empty his account. Thanks Microsoft I knew you where so helpful to hackers while making my life shittier all the while.

It's just looking in a sqlite file and listing the jpeg directory. The only extra step is running icacls to let the user read the files.

I saw someone on mastodon say something like, "don't tell your IT department not to use recall to protect employee or customer data. Tell your legal department that all your recall data can be subpoenaed for discovery."

They encrypt the damn start menu and they cannot encrypt this?

Likely because there was too much CPU overhead decrypting and having the LLM query the Recall image database all dynamically

It requires full disk encryption doesn't it? If someone already has access to your account then they can access this data the same way you can. The new issue here is that this silos a load of private data in one easy to grab location. Users would have to set up the filters perfectly to prevent recall capturing anything more sensitive than what's already accessible to their account. This is in a world where many users are probably storing their passwords in a Word document on the desktop.

It could be that anything you encrypt has to have its encryption key in some place inaccessible to these same hacker tools. If your computer uses Bitlocker, for instance, you need to enter a 6-digit code each time you turn it on.

Best guess, they had such a high expectation of "convenience" for this feature that they couldn't justify any kind of security key. Which is still a dumb explanation, obviously.

not caring ~~enough~~

So... how does the user actually ®ecall™ anything? Do they have to ask M$ Co-pilot™ AI to get it from The Azu®e™ Cloud? Because I'm pretty sure a hacker could do that just as easily.