Heyo,
First off, sorry to hear that this has happened to you. It's never the for tools to be abused like this and for support be to be so unreachable.
I've gone ahead and sent you an email so I can get some details and we can work to recover at least some of the content.
Q.
Disc: Googler, but I don't work in YouTube.
:)
I hope OP got in touch. Thanks for doing the thing.
Reading the thread made me think of a curious idea. It's got a couple layers of nuance, but could ultimately be really cool.
In short: have a proactively-updated list somewhere that shows the presentation and content of every single email that will be sent before someone is due to lose access to a given service or their whole account.
For example this could look like a (perhaps loosely-interconnected) web of support.google URLs with content like "you might receive an email like the following" <insert HTML extract/embed or screenshot of exactly what the email looks like here> "and you can do" <insert brick wall here :)>.
If this were a reliable resource that stuck around, I could totally see a community-maintained tool emerging that would monitor for these sorts of emails then eg send emergency SMS/notifications/etc so that carefully-coordinated action could be undertaken immediately by competent users. These notifications could be sent to a tech-savvy family member or associate/colleague, who likely would not otherwise have access to the account in question. This person could then gain impromptu access (perhaps via remote access) and take immediate action such as generating takeouts (and perhaps sharding them to raise the chances they complete in time), individully downloading or screenshotting critical info, and firing out bulk alerts reminding of alternate contact info. Basically disaster-handling to try to end-cap workflow disruption while the mess (hOpEfUlLy) gets fixed.
I can't help but think there's probably a disconcertingly large proportion of grandmas and effectively-illiterate users being caught in situations where they have very literally no idea how to proceed (eg, none of their Facebook friends are going to have a clue).
It would seem there is enough awareness and reactivity localized around HN to make a difference to that 99th percentile, if harnessed in a constructive and non-disruptive way.
It's a nuanced idea, but I'm reminded of the old YouTube story where optimizing the homepage made the site seem to load slower because it became viewable by a large demographic of people using sub-2G connections. Here, I can certainly see a temporary uptick in apparent legitimate account closures as the true false positive gap is surfaced, and that might translate to a bit of a suboptimal reputation hit. But in the ideal case, the new feedback would be integrated quickly, that transient spike would soon be forgotten, and in the long term, integrating the new data would help to further micro-optimize the performance of the 99th percentiles involved with the ML analytics processes in ways that exceed expectations.
I'm thinking of the bigger-bigger picture; I understand it's simply impossible for Google to provide tier-one support to every single one of their four billion+ free users. This approach could implement an unorthodox workaround that lets motivated individuals contribute toward closing the loop of false positives. Currently it is effectively impossible for Google's userbase to take charge of this particular edge case. I am sure the number of people who would be very happy to take preventative action against false positives completely outweighs (perhaps by an order of magnitude) those who posture and rage and make themselves look stupid in the forums when things don't go their way.
Knowing what all (or almost all) warning emails look like ahead of time is the only way a significant dent could be made in the false positive statistics; there's nothing stopping a detection project from getting off the ground today, but the lists of what content patterns to look for would have to be forged in blood by users who've had their accounts shut down, with this status quo compounded every time the wording of the warning emails changes.
The one caveat emptor I can think of is that there is of course no way to prevent malicious users from taking advantage of a community initiative that enables a heightened baseline of proactivity, especially if it goes viral and is easy to use. But I'm honestly curious... how might this be the end of the world? The worst the malicious users can do is download their data, which a) they probably have a copy of anyway, and if they don't, b) wouldn't be a great look in legal cases where a defense can say "here is evidence ABC used XYZ software to monitor their account for shutdown, which is why shortly after *point* this email was sent they *point* submitted a takeout request 5 minutes later."
I can definitely see some legal/regulatory/policy hesitancy (sigh), but (from my position as a relatively naive random internet user) I'm hopeful that hesitancy would mostly be without precedential substance, and this could actually viably happen with a bit of gruntwork and convincing.
It's not a terrible idea, and personally, I believe we should be doing more proactive reach-out when events like this are happening. Thing like text messages, push notifications, and big red scary banners on the home pages of unrelated products.
I'll file a Feature Request internally, but given the scope I can't imagine things will be too quick.
Disc: Googler, not in YouTube
If making changes to Google-side reach-out is actually possible, that's absolutely the far more realistic scenario here. My previous post is more of a Rube-Goldberg overengineered attempt at making the hockey puck meet the stick rather than the other way around.
Side-channeling "here's what our emails look like" kinda feels like kicking the info out a side door or something, which I can totally see internal clearances and perhaps legal just not liking the look of... and even if that approach did succeed, the less-straightforward internal handling associated with such an indirect strategy may well incur delays in updates and perhaps even occasional accidental article deletion (where the replacement articles wind up at different URLs and people have to take ages to find them... yep, Rube-Goldberg machine).
I can imagine a new broad scope in myaccount.google for "enhanced notifications" "if there is any suspicious activity with my account", with an associated stronger connection between account flagging events being classified in with suspicious activity in general.
Perhaps I'd first add other users' information under People & Sharing, allow them access to the relevant info in Data & privacy, then opt them in to security notifications in Security. (Where do I send my PM application again? ;) )
I'd probably want to be able to add direct SMS numbers (handled "magically" or "hands-off", like one-time verification codes etc, with the exception of being stored long-term) so that in worst-case situations if an entire group of accounts gets super-nuked or whatever (which presumably blacklists the contact methods associated with those accounts, including phone numbers - hence the idea of magical handling) I still receive last-resort notifications in case of edge-case mistakes. Obviously third-party email addresses would be ideal to add too (with the requisite amount of confirmation bustlework - I'm reminded of the absolute tantrum Gmail very appropriately pulls when you enable forwarding, maybe that would be too strong here but I imagine it might make for good inspiration). Adding Google accounts to send FCM push notifications to multiple stakeholders could also be a good idea (and, FWIW, may also create a helpful source of high-signal data to contribute to cross-account security analyses processes).
Generally I'm trying to cover 2-3 of broad bases here: that of having last-resort notifications still function in scenarios where the AI mistakenly decides everything about an account (or worse, set of accounts) is worst-case-scenario not-ham; making it possible to send those notifications to multiple users/contacts, including contacts that do not typically have access to a particular account; and making sure the notifications and pings actually get through even when the system really doesn't like someone.
Big scary banners on unrelated products is a great idea! I would never have thought of trying to vie for something like that, would have thought it would be possible.
Hmm, could you modify the user account avatar service so that if the user is requesting their own avatar they get served a giant exclamation instead? That could help serve as a perpetual reminder that would show up *everywhere* - Google homepage, account icons across the web and in random Android apps, etc etc. (Potential blocking issue: things might cache the avatar and get stuck.)
I've also occasionally seen how the homepage occasionally shows a very small infobox with a bit of text in it toward the lower quarter of the page. Perhaps this could be hijacked to prioritize showing the scary warning as well.
Totally understand that this'll move at "eventual consistency" pace :)
Feel free to copy the text of this and my parent (dumpster-fire...) comment into the feature request if that's helpful.
My motivation here is mostly a reaction to low-grade obsessive paranoia about AI glitching out on my or my family's accounts, and a strong interest in doing whatever I can to mitigate the fallout in the worst case scenario.
Thanks very much for replying!