It's been a hectic week for Bluesky Trust & Safety. With the Brazilian wave, reports have skyrocketed, bad actors targeted the site, systems needed adaptation, and new harms emerged. 1/7
Comments
Log in with your Bluesky account to leave a comment
Phenomenal work, kudos to the entire team, and thank you for your devotion to the platform. This was a tremendous amount of stuff piled onto the team's weekly workload:
Thank you all for your work, and for keeping us in the picture. And keeping each other safe can become a shared, community responsibilty, as the culture and ethos of Bluesky becomes established with so many more members. :)
We talked about Sec. 230 in my privacy law course this week, and I mentioned the steep challenges T&S teams face, referencing Bluesky's recent explosion of growth. Great work.
You guys are incredible... can't believe the amount of workload you all must have taken on since the migration and you're still going beyond your limits to try to get everything sorted asap besides rolling out all the new things, we feel lucky and safe being on here, thanks for so much hard work! 💜
You are bad actors yourself when you apply erratic labels to people without even give a notification, let alone react to objections.
You personally fucked up big time at Twitters , and you will fail here
I might never go back to the fallen X… this seems to be a much more pleasant place to share ideas, have great conversations, a avoid hate speech… so glad I’m here…
Yeah! If it happens you can just block the person, they will never bother you and your circle of people that engaged in the posts (they won't see the blocked person by what I've understood)
Plus there're lists to massively block unpleasant ppl 🙌🏻
"While bad faith actors exist and context matters, most cases involve two people trying to spend time online."
"multiple bad actors targeted the site."
I like this place, but like, you gotta realize how a 9-day difference between these threads feels obtuse, right? "Most cases" is sorta relative.
Hi Aaron, the Brazilian market is excited about the possibility to substitute X with Bluesky. Next week is Rock in Rio. Please see separate e-mail. Thanks best Alex
I have faith in you and the few members of your team. You have been consistently doing the right thing and I know you will continue in that vein. Thank you for the update and thank you and your team for being here!
Typically, we receive about 20k reports weekly, mostly in English and Japanese. In the last 7 days, we've seen over 270k reports, predominantly in Portuguese. This surge has stretched our resources thin, despite mods and myself working continually to address the most severe harms. 2/7
Adding to the challenge, multiple bad actors targeted the site. One group, similar to 4chan users, conducted raids targeting black, trans, and LGBT+ users. Our takedowns have increased 2.5x from an average week. Please report abusive accounts/campaigns - we'll address them ASAP. 3/7
This seems like an inevitability with any social media platform in general. Once a big popularity tide shows up, that exponentially rises the good — and the bad. Proof that technology is not the one to blame per se. And fixing the human condition isn't a matter of code. It never is.
I wonder if the cat might already be out of the bag, but a kind of social credit system.
I liked the initial invite only stage very much for that reason. Attaching some degree of responsibility to the issuer would have been the obvious next step.
Calling all sociologists!
CSAM content has risen 10x week-over-week. While we detect confirmed cases quickly, manual review is still necessary for NCMEC reports. This increase impacts our mods, who are repeatedly exposed to such content. 4/7
Though I am sure the challenges are vast, how does crowd sourcing perform in these scenarios? I am sure many people would be very happy to assist in getting that out of the community.
I would like video to only be available to accounts that qualify in some way. Age of account plus number of posts or something. Achieve that without an intervention against you and it unlocks for you.
That's what "confirmed cases" are, I believe: photos that match what's already known and thus in the software.
*New* material may get flagged, but considering that even the best automated systems are full of false positives... The mod team has to use human eyes to winnow the trauma bombs. 😔
Yes, "confirmed cases" = hash match from PhotoDNA. There's a frustrating legal issue where some courts have found that just a hash match isn't enough for a search warrant: a human has to have verified that it wasn't a false collision, and more frustrating issues with NCMEC being a government actor.
Our automations were designed for activity levels prior to the 2.5 million new users. Rules that previously worked now generated hundreds of false positives as a huge proportion of the network is suddenly new accounts under a week old. 5/7
The influx has likely shifted Bluesky's demographic composition. While we don't collect specific data, conversations suggest a younger crowd joining. This brings new challenges, like increased discussions around eating disorders. 6/7
Re: demo shifts, doesn’t this include bad actors from your 3/7 and 4/7 posts?
IMO, auto-mod and block features isn’t enough. Bad actors shouldn’t be platformed period. Bans would be more effective at stopping the proliferation of hateful ideologies and abuse of marginalized people.
Please consult subject matter experts (folks expert in both eating disorders and social media) before making policies around discussion of eating disorders.
We've adapted for now, but there's a clear need to update our community guidelines and develop better policies to address these emerging harms and limit the promotion of potentially harmful content. Your patience and support during this period of rapid growth are greatly appreciated. 7/7
It might be out of budget, but check out https://safer.io - they have AI tools that can help, but also moderation tools (like using a flashlight in the dark when verifying, instead of seeing the full image) which may help your mod team
Please consider implementing a labeling service that integrates with identity verification services like https://openpassport.app, https://www.proof.com, etc. to issue verifiable credentials on the DID. Then custom feeds can be built that only show content from these accounts.
Thank you so much for defending us and making this a safe space for POC and LGBTQ+ users! It feels so refreshing to actually be seen by a social media platform as people and not just statistics and numbers that can be ignored! Thank you!
Maybe if the moderation team is overwhelmed with reports that fall short of clear hate speech, the label is put on. Perhaps we need to look through and find the worst examples and report only those.
Live moderator exposure to hateful content is no joke. These folks will need resources, which I’m sure you’re aware of.
We’ll one day arrive at the moment when especially automated and algorithmically hopped hate speech will get treated legally as assault. Sooner the better.
Your work is very good.
Here bad contents were deleted quickly after our complaints.
I think could be a good idea create a new kind of complaint: "fake news". Because there are groups of no good people in Brazil using internet for sharing fake news with political intentions.
Nice work!
You are starting to operate at a time before the elections, the demand will increase even more and I believe it will reach 500k to one million per week close to the election period
We Brazilians are a "slightly" messy bunch. I apologize for that and please be a little more patient with us. Oh! Of course, there can be no tolerance for hate speech.
We here in Brazil absolutely love your care in keeping this network healthy. A message sent by BSKY went viral in which they advised a person who might have suicidal tendencies. Your care surprised us. Keep up the great work!
Thank you for your work! We brazilians ended up here kind of by chance but I personally want to stay, even if X comes back. This place makes me feel a lot more at home.
Appreciate you and the team! Please pass along that the community is grateful for all the work and sacrifice! That sounds so challenging, and I’m sure I’m not the only one who’s proud of y’all!
That just leads to abuse of privilege and soft bigotry. Exclusivity is bad actually. I am shocked they were not ready to scale, but gkad they understand the need to.
Social media isn't a clubhouse. It never will be. It's a privately run public square. That allows for limiting access to those who wish to cause harm, but that's about it. Usually when people talk about undesireables they do not just mean open bigots.
You are mistaking forums for social media. No, there are not. Not successful ones, anyway. Yes, it's a privately owned public square. Anyone can see and respond to you, follow you, etc. The primary use of sm is info seeking and sharing. It's been studied.
It has, but at some point it just isn't going to work. The more people that come, the more invites you generate. And even if one bad person gets a stray invite, they can now start inviting every other of their friends into it.
And this is supposed to be open, not just a walled garden.
yeah but its easy to prune an entire branch of bad actors if a person was giving invites to terrible people, since the invites could be traced back to the source
Probably should have email verification, but then that would make it harder for anti-LGBTQ+ and other Nazi variants to rapidly create sock puppet accounts, and we know you’re really proud of your statement where you said their presence and opinions matter as much as ours!
Appreciate the grind that goes on under the surface. Thank you for a platform I am increasingly treasuring (in the oily toxic wake of X-Twitter, this place was an oasis.
Good job! I crave this kind of social network and was in a toxic relationship with Xwitter. Bluesky is treating me so nicely, in a way I'd forgotten I deserved.
What about sexual content not being hidden when the option is enabled? For example, if you go to #brasil tag the majority of the content are gay things. The images are not always hodden and the text is not hidde. I dont even want to see it. I dont want to see the entire post not only hide the image.
This is disgusting. A lot of gay dicks flooding the feed. And this is not about gay, it is about sexual content no being filtered but only images being hidden. You are apending bandwith and screen space with things that the user flagged as not interested.
dude - you can literally mute hashtags you don't want to see on your timeline ... just filter out the stuff you appear to be a tad "uncomfortable" with
I dont want to mute brasil. If there is a flag to hide adult content it shoul hide the entire post not only the media. That is the point. The flag does not remove the pos, only hides the image... I can still see the post, avatar and the message...
All the posts that you have issues with have multiple hashtags. Just mute the ones you don't want to see or just block the accounts as the majority of the posts are coming from just a few accounts.
Oh so the reason you've been able to take no action against the Nazis and transphobes who have been on here for weeks/months is because this week has been busy? Unblock Jane and the other people who have been spoon feeding your baby brained team and give them the conversation they deserve.
"You can call on the community for assistance, allow requests for analysis of block lists, and advise users to send links and screenshots in reports of misconduct. This could save you some work.".
Comments
Thanks for all that you do.
I hope you have everything in your life just leave you, and when you have nothing, the Nazis you're coddling turn on you.
...which in a way is kind of a good thing.
You personally fucked up big time at Twitters , and you will fail here
Plus there're lists to massively block unpleasant ppl 🙌🏻
"multiple bad actors targeted the site."
I like this place, but like, you gotta realize how a 9-day difference between these threads feels obtuse, right? "Most cases" is sorta relative.
Let's keep on (we) tagging and (you) bagging those extremists!
I would hope they've already been reported.
I haven't checked them myself yet.
https://bsky.app/profile/brainnotonyet.bsky.social/lists/3jvplg6i7vb2p
But seriously, the team did and is doing an amazing job.
I liked the initial invite only stage very much for that reason. Attaching some degree of responsibility to the issuer would have been the obvious next step.
Calling all sociologists!
*New* material may get flagged, but considering that even the best automated systems are full of false positives... The mod team has to use human eyes to winnow the trauma bombs. 😔
IMO, auto-mod and block features isn’t enough. Bad actors shouldn’t be platformed period. Bans would be more effective at stopping the proliferation of hateful ideologies and abuse of marginalized people.
You don't need to reinvent the wheel.
It might have to be a case of:
- label applying
- removing the warn/badge options (for off and hide).
If the users aren't visible and are hidden, the reports after labelling will reduce, am I wrong? 🤔
I was subjected to mislabelling and as a result the attacks on me including users telling me to "off myself" increased...
I have several suggestions further down the article.
Mind my writing tone. I was annoyed.
https://amicite.co.uk/when-bluesky-becomes-redsky-mislabelling
Really appreciate the work it takes to keep the site clean. Especially with so many new users.
Is there a reason why it’s been labeled and not removed?
https://bsky.app/profile/naomicunningham.bsky.social
We appreciate your hard work!
We’ll one day arrive at the moment when especially automated and algorithmically hopped hate speech will get treated legally as assault. Sooner the better.
Here bad contents were deleted quickly after our complaints.
I think could be a good idea create a new kind of complaint: "fake news". Because there are groups of no good people in Brazil using internet for sharing fake news with political intentions.
Nice work!
And this is supposed to be open, not just a walled garden.
Most of Brazil would still be waiting for invite codes.
We appreciate it
"You can call on the community for assistance, allow requests for analysis of block lists, and advise users to send links and screenshots in reports of misconduct. This could save you some work.".
you are very much appreciated