Bluesky, the decentralized social network and frontrunner alternative to Twitter, has been hailed as a wonderland of funny posts and good vibes. But a moderation policy change that followed a death threat against a Black user has many on Bluesky questioning if the platform is safe for marginalized communities after all.
Bluesky had around 50,000 users by the end of April. Its users have doubled since then, and as it gains more, it also faces increased pressure to crack down on hate speech and other violent comments. As a soon to be federated platform, Bluesky is at a turning point that could set the precedent for moderating decentralized social networks. Though robust moderation wasn’t one of Bluesky’s founding principles, many users expect the site to be more proactive in refusing platform bigotry — even if it conflicts with Bluesky’s decentralized goals.
Bluesky has not announced a specific timeline for federation. In a May 5 blog post, the Bluesky team said it plans to launch a “sandbox environment” to begin the testing phase of federation “soon.”
It started in the hellthread last month. The thread, which initially formed when a coding bug notified every single user in the thread every time another user responded to it, grew into a chaotic, seemingly infinite discussion board with countless subthreads. Initially a shitposting outlet, the thread has devolved into a hotbed of discourse — opening the door for rampant racism.
Aveta, a software engineer who has invited hundreds of Black users to Bluesky in hopes of recreating Black Twitter, replied in the thread asking people to stop posting R. Kelly memes. Aveta is well known on Bluesky for expanding the Black community on the platform, and is an outspoken advocate for acknowledging Black influence on internet culture.
Last month, she had a dispute with Alice, a Bluesky user who went by cererean, over comments that Alice made about the growing Black community. Alice made multiple racist posts in the past month, including one that said that Black users are welcome to create their own spaces if they don’t want to be somewhere that “reflects the demographics of the Anglosphere.”
In response to another comment about Aveta’s hellthread interactions, Alice suggested that Aveta get shoved off “somewhere real high.” Aveta, who declined to comment out of fear of harassment, described Alice’s comment as a death threat in posts on Bluesky.
Other users reported Alice’s comment as a violation of Bluesky’s policy prohibiting extreme violence. Bluesky’s moderation team did not initially ban Alice, and invoked further outrage among users when Bluesky CEO Jay Graber announced a change in the platform’s policies that appeared to excuse comments like Alice’s.
“We do not condone death threats and will continue to remove accounts when we believe their posts represent targeted harassment or a credible threat of violence. But not all heated language crosses the line into a death threat,” Graber said in a weekend thread. “Wisely or not, many people use violent imagery when they’re arguing or venting. We debated whether a “death threat” needs to be specific and direct in order to cause harm, and what it would mean for people’s ability to engage in heated discussions on Bluesky if we prohibited this kind of speech.”
Under Bluesky’s new policy, any post that threatens violence or physical harm — whether literal or metaphorical — will result in a temporary account suspension. Repeat offenders will be banned from Bluesky’s server, but once Bluesky finishes the “work required for federation,” Graber said, users will be able to move to a new server with their mutuals and other data intact.
Like Mastodon, Bluesky aims to be a decentralized, federated social network. It isn’t federated yet, so all users still interact on Bluesky’s server and have to abide by Bluesky’s policies. Once it is federated, any users on any server on AT Protocol will be able to “opt in” to a community labeling system that would include certain content filters.
That means that under Bluesky’s new content moderation policy, a user who was suspended for hate speech or making violent threats would still be able to engage with other servers running on AT Protocol. Bluesky has always been transparent about becoming a decentralized social network, but the swift action it previously took against users who threatened others convinced many Bluesky early adopters that the platform would continue to shut down violent or hateful rhetoric.
“While this may not be your vision necessarily, I think a lot of people are less concerned with moving to a new instance of Bluesky, than making sure bigots are not able to have *any* instance on here,” Ben Perry, a Bluesky user also known as tedcruznipples, replied to Graber’s thread. “They shouldn’t be given the opportunity to have federation and proliferate their message.”
Bluesky rolled out custom algorithms the day after Graber announced the new moderation policy. The feature allows users to choose from Bluesky’s “marketplace of algorithms” instead of just seeing content from the “master algorithm” that most social media sites employ. Like Twitter lists, users will be able to toggle between the “What’s hot” tab, a tab of people they follow and tabs for custom feeds they’ve pinned. The “Cat Pics” feed shows, predictably, cat pics, while other feeds lean more toward memes and NSFW content.
But many Bluesky users — particularly Black Bluesky users — questioned the timing of the roll out. Rudy Fraser, who created a custom algorithm for Black users called Blacksky, said it was unfortunate that Bluesky tried to offer custom algorithms as a “solution” to the moderation debate.
“As if a new feature would resolve the underlying issue and as if they couldn’t just ban the offending user,” Fraser said. “Some form of programmable moderation is on the horizon, but there’s not yet a prototype to see how it would work… There are already ways around the NSFW tags for example. They need to find those bugs before they reach critical mass.”
Moderating decentralized social networks is a challenge that by definition offers no easy solutions. In the case of Bluesky, establishing and enforcing centralized community guidelines for all users seems antithetical to Bluesky’s aspirational system of federation and customizable moderation. On Mastodon, moderation is unique to each server, and one server’s policies can’t be enforced on another server with a different set of rules. To be listed under Mastodon’s server picker, servers must commit to the Mastodon Server Covenant, which requires “active moderation against racism, sexism, homophobia and transphobia.” While most prominent servers abide by the Covenant, unlisted servers aren’t held to a minimum standard of moderation.
The fediverse, a portmanteau of “federated universe,” promises a vast social network that can exist beyond the authority of a single institution. Though there are benefits to that level of independence, the approach to community-led moderation is often optimistic at best, and negligent at worst. Platforms can absolve themselves of the burden of moderation — which is labor intensive, costly and always divisive — by letting users take the wheel instead.
Allowing communities to moderate themselves also allows violent hate to go unchecked. In a recent skeet, software developer Dare Obasanjo pointed out that many “techno-optimistic” approaches to content moderation fail to account for context.
“A user with a virulent racist history wishing harm on a BIPOC is different from the same comment in a debate about MCU versus DCEU movies from otherwise well behaved users,” Obasanjo wrote. “A legalistic discussion of whether ‘someone should push you off of a tall building’ is a ban worthy offense misses the point completely. The question is whether you tolerate openly racist people wishing harm on BIPOC on your app or not?”
Bluesky employs automated filtering to weed out illegal content and do a first pass of labeling “objectionable material,” as described in a blog post about the platform’s composable moderation. Then, Bluesky applies server-level filters that allow users to hide, warn, or show content that may be explicit or offensive. Bluesky plans to let users opt-in to certain filters to further customize their individual feeds. The ACLU, for example, can label certain accounts and posts as “hate-speech.” Other users will be able to subscribe to the ACLU’s content filter to mute content.
Graber wrote that the layered, customizable moderation system aims to “prioritize user safety while giving people more control.” The company hasn’t publicly clarified whether or not it plans to hire human moderators as well.
Moderation in Bluesky’s early days has been met with mixed reception from users. In April, Bluesky banned a user who went by Hannah for responding to Matt Yglesias with, “WE ARE GOING TO BEAT YOU WITH HAMMERS.” Many Bluesky users protested the ban, and insisted that the user was joking. Days later, Bluesky swiftly banned another account that had harassed other users with transphobic comments.
Hannah’s hammer ban resurfaced on Bluesky in wake of the moderation policy change. Black Bluesky users questioned why the threat against Aveta wasn’t taken as seriously as Hannah’s reply. Mekka Okereke, director of engineering at Google Play, described Hannah’s comment as “metaphorical and inappropriate,” but pointed out that “people just can’t empathize when Black women are the subject.”
“And as I’ve said on here before, ‘echo chamber’ is a specific term mostly used by right wing news outlets to describe any place that tries to make Black, brown, and LGBTQIA people feel safe,” Okereke said in a post. “And the ‘truth matters’ philosophical pedantism only seems to come out when we’re talking about making Black women feel safe online.”
Pariss Athena, who founded the job board Black Tech Pipeline, conceded that no online space is truly safe, but pointed out that direct racism, transphobia and anti-Blackness are “not blurred lines.”
“Accountability and permanent action needs to be taken the way that they are offline,” she wrote in a post. “What’s odd is that these things have already happened on Bluesky but it seems to move much slower when it comes to Black ppl.”
Users once praised the platform for refusing to harbor hate speech — a decision that seems hollow to many Black Bluesky users after the most recent moderation policy change. Bluesky set itself apart from Twitter not only for its customizable features, but also for what appeared to be a serious interest in protecting marginalized groups. The recent policy changes, however, leave many doubtful that Bluesky can maintain that safety.
Fraser isn’t sure that Bluesky is ready to be “proper stewards of a safe space for marginalized groups,” but is determined to stay on the platform for the time being.
“I am genuinely optimistic for the protocol’s potential and hope they will make a clearer effort to build alongside marginalized communities going forward,” Fraser said. “I’m a firm believer in ‘Nothing about us without us.’”