reduce legacy spam #3322
Description
Starting a brainstorm for some action we could take to reduce legacy spam, which consumes a decent amount of our support time. We're facing multiple vectors that have become worse with folks using ai agents to generate contextual reasonable-sounding stuff that gets past most of our spam filters and rate limits etc -- spam comments, spam users, spam pubs, and spam uploads to assets.
A few things come to mind:
- add a captcha to user signup (see feat: add honeypots and (mostly) invisible PoW captchas to highly spam prone forms #3488)
- add a captcha to new discussion creation (or a honeytrap -- there are some new techniques for llms that are interesting) (see feat: add honeypots and (mostly) invisible PoW captchas to highly spam prone forms #3488)
- give communities ability to turn of discussions (see feat: community creation spam block #3479)
- prepend assets uploads with pub or community id (to make it easier to track down and ban existing perpetrators) (see feat: include context in uploads #3409 and feat: be able to tag users as spammy, add user superadmin dash #3472)
- make html uploads download instead of render inline (see feat: only allow iframes to render html #3471)