The presumption of innocence is for courts of law.
The court of public opinion has no such rules.
If we, the people, see a credible accusation against someone who seems like the type of person who might be guilty of it, we are free to call it like we see it.
Naturally, this can lead to problems: after all, who is "credible" and who "seems like they might be guilty" can be (and very much have been) subject to a lot of bigotry. This is why we have the rules we do in courts of law. But anyone trying to enforce actual legal court rules on public online discussions generally just ends up looking like a jerk.
at uncloseai dot com we took up the AGPLv3 only because it will be maximum freedom if anyone like Anthropic or OpenAI get caught training on us they should open their weights to satisfy the license and terms of use. we have a browser extension & a white paper also covered by the same license I will also settle out of court to maximize user freedom by opening the model weights & distillation for free.
my favorite part of lotus notes is the password scrambler that prevents people snooping over your shoulder to steal password, I implemented this on my front-end at unsandbox dot com you can test it on teh console page even without an account, the portal itself does not have accounts.
