Why haven't you implemented an AI spam filter?

Question

Why haven't you implemented an AI spam filter?

Allan West 25

I get a lot of spam and phishing. I report them. Add them to my blocked list, etc. But they just keep coming. A child could tell they're phishing emails. My dog knows they're not legit. I don't have to open them or look at them for more than a second to know they're spam. But MICROSOFT can't seem to figure out a way to stop them filling up my inbox. Is that just beyond the capabilities of your AI? Shouldn't this be a lot simpler than image analysis? It's text. I'm just astounded. Have you seen what everybody else is doing with AI!? If I copy and paste one into Copilot, I bet it'll make a pretty good guess at what's spam and what isn't. Even Bing will probably get it right. AND BING SUCKS.

I saw somebody else posted a similar question on here, and your answer seemed to be "da guy sending the spam keep changing his email address. nothin' we can do." Is that all it takes to outsmart you guys? Changing your email address? Did you cryogenically freeze your tech-support team in 1998 and you've just unfrozen them now? None of this is rhetorical, I'm genuinely interested in the answers. I DO NOT UNDERSTAND HOW THIS IS POSSIBLE.

You also threw in "if you don't want to get spam, maybe you should filter out those keywords".

Nobody 'wants' to get spam, unless they're researching spam trends, which clearly nobody at Microsoft does
That doesn't work.

BAR 15 Reputation points

2025-10-27T19:58:26.87+00:00

I find it particularly annoying how many messages in my outlook account are phishing e-mails from places pretending to be Microsoft.

To add to the frustration, when you attempt to create filter rules, the webmail interface doesn't let you see the domain it came from, just what the sender put in their header. MS's efforts to make the inbox look sleek and modern have, unfortunately, broken an important tool in spam detection by the user; the ability to see at a glance whether the domain matches the expected domain of the sender.

Honestly, if MS simply allowed us to do things like "delete all e-mails that have emoji in the subject line" or "delete all e-mails from domains that include cyrillic text" or any of a variety of other rules (if "from" contains fargo but sender domain does not end with fargo.com, for example) then at least we'd be empowered to fix this dumpster fire ourselves but, unfortunately, the interface for creating rules seems intentionally obtuse and useless.

Answer accepted by question author

Vasil Michev 127.9K MVP Volunteer Moderator

I share your frustration, and often ask the same questions as the ones you listed above. I remember days where my Exchange Online mailbox was happily sitting on 0 junk email, but for the past few years things have turned around, and nowadays I'm happy if the number of junked messages is below 100.

Some of this is definitely on Microsoft, and my explanation for the likely reason behind it is that they are simply being over-cautions. With ExO being a commercial service and powering email for 99% of the large enterprises out there, the downsides of letting obvious (but often harmless) spam messages pass outweigh the risks of blocking a potentially important message from reaching someone's work mailbox. Thus we end up with random "girls" asking what my plans are for the weekend :)

Phishing is a different story, if you constantly find such messages hitting your Inbox, you should review your hygiene settings and switch to something stricter, where possible. And you should definitely keep reporting those, as unfortunately there is also the "evolving" attacker element. There is no perfect solution, but in my experience Microsoft does a much better job handling phishing emails compared to plain worthless spam.

And you are correct that blocking keywords, or even addresses does not work. It might be a solution for notification-like emails from providers that do not respect unsubscribe requests, but they can easily change the sender address. What does work is blocking domains and IP ranges for repeated offenders. It wont solve all such issues, but can help reduce them. Here's the relevant documentation: https://learn.microsoft.com/en-us/defender-office-365/tenant-allow-block-list-email-spoof-configure#use-the-microsoft-defender-portal-to-create-block-entries-for-domains-and-email-addresses-in-the-tenant-allowblock-list

0 comments

1 additional answer

Your answer

BAR 15 Reputation points

2025-10-27T19:58:26.87+00:00

I find it particularly annoying how many messages in my outlook account are phishing e-mails from places pretending to be Microsoft.

To add to the frustration, when you attempt to create filter rules, the webmail interface doesn't let you see the domain it came from, just what the sender put in their header. MS's efforts to make the inbox look sleek and modern have, unfortunately, broken an important tool in spam detection by the user; the ability to see at a glance whether the domain matches the expected domain of the sender.

Honestly, if MS simply allowed us to do things like "delete all e-mails that have emoji in the subject line" or "delete all e-mails from domains that include cyrillic text" or any of a variety of other rules (if "from" contains fargo but sender domain does not end with fargo.com, for example) then at least we'd be empowered to fix this dumpster fire ourselves but, unfortunately, the interface for creating rules seems intentionally obtuse and useless.

Answer 1

Alex Dunstan 0

You know outlook is broken when you need to use power shell to have a basic experience to rid spam. the company who is constantly watching you can’t even pattern recognise spam effectively.

0 comments