Yes, access to production database is fairly common (for certain job functions, at least). Unaudited and unfettered database access is much less common. Sure, it happens, but it is rare - especially for something at the scale (& attractiveness to hackers) of Instagram. And yes, an audit trail doesn’t mean your manager will be immediately alerted, and there are people who won’t think of the audit trail and go snooping in prod anyway - so it is possible, but I just don’t think it’s very probable ¯\_(ツ)_/¯

And a moderation tool for direct messages? Which are E2E encrypted? That doesn’t make much sense to me. What moderation function would a “list of people they have DMed in the last 2 years” serve? I guess it could be used to determine if somebody has been harassing someone else - but the block feature exists, why would it reach a moderator in the first place?

and frankly you should assume that this is happening behind the scenes at every company.

Look, I operate under the principle of “anything that I put online, will be eventually public and linked to me” (which is why I would never answer the original question, even with an anonymous account that isn’t linked to my email) and “everybody sucks at infosec” - but that doesn’t mean Instagram employees have a handy way to access a human readable list of people I have DMed.

Occam’s razor is in favour of the girlfriend getting the info the old fashioned way - snooping on the OP’s phone

Ya, I mean Instagram is no bastion of privacy, I’m sure - but most managers wouldn’t be thrilled to learn their employees were accessing the production database for fun. It’s less a “but you violated our customer’s trust” and more a “you idiot, why you tempting fate, we are generally one typo away from the whole thing crumbling down anyway!”. And surely no company bothered to build a nice tool that’ll let their employees peruse the DM list of a random user - we can barely get them to build us actual monitoring infrastructure till something breaks! So one would have to put in some effort into gathering this information. Running background checks for some random friend - the risks and effort doesn’t feel like it would be worth it. It seems more likely the girlfriend peeked at OP’s Instagram client herself, or just took a guess, and made up “a source working at Instagram” as a plausible excuse.

You’re on the lemmy.world instance, so you can reach the admins by emailing info@lemmy.world, or posting in the support forum !support@lemmy.world

Now to answer whether there’s a difference between being promoted and doing it yourself - In this case, it’s suspected that session tokens were compromised. You know how when you enter some events, they vet you/your ticket once at the door and then put a stamp on your hand? If you go out and want to get back in, you don’t have to do the whole verification song and dance again, just show them your stamp? Well, that’s pretty much what a session token is - Lemmy vets your password once when you log in, and gives an unique session token to whatever browser or app you used to log in. That way, when you reopen Lemmy, you don’t have to enter your password again.

Now that token is compromised, you have to assume a hacker has your unique token. When you logged yourself back in, Lemmy did the whole validation process again and gave your browser/app a new, unique session token - that’s just how logging in works. But the important question is, did it invalidate the old session token when you logged out? Otherwise the hacker can still show the old token and pretend to be you.

Now if your browser/app prompted you to log-in today, you can be sure that your browser/app tried to get into Lemmy and was denied access. That means you can be sure your old stamp/token is now invalid. Logging out and in yourself doesn’t give you the same guarantee - you will have to check Lemmy code (or run some experiments) to know if logout does actually invalidate the old token. I haven’t validated Lemmy’s code, but I will say most half decent software will invalidate your token when you log out. If you want an extra layer of protection, change your password as well - even the software devs that forget to invalidate tokens on logout usually remember to invalidate them on password changes.