Musk’s announcement that he will grant a “general amnesty” to banned Twitter accounts starting this week comes roughly a month after he took over day-to-day management of Twitter and laid off nearly half of its workforce—including the entirety of the human rights team and an untold number of outsourced content moderators who tracked hate and harmful posts on the site. Key security, privacy, and trust and safety executives have also left the company since Musk took over.
“Musk’s takeover of Twitter is nothing less than apocalyptic for our communities,” said Thenmozhi Soundararajan, executive director of Dalit civil rights organization Equality Labs. “This is not about hurt feelings. This is about direct calls to violence.”
Soundararajan took part in a press call on Monday in which representatives from a number of advocacy groups spoke out, insisting that reinstating the accounts of known inciters of violence would pose a threat to human rights worldwide. The speakers highlighted how Musk’s decision would have an outsize impact on regions of the world outside the U.S., where Twitter has either been integral to the safety and security of vulnerable populations or has contributed to political and social unrest.
“For those of us in the global majority, our Twitter experience does not look anything like that of users in the United States,” said Rosemary Ajayi, lead researcher at Digital Africa Research Lab. “In Nigeria…we have bad actors, including those engaged in coordinated inauthentic behavior, and others evading suspension who have been repeatedly verified even after their accounts have been flagged to Twitter.”
The activists’ fears aren’t unfounded, according to a study conducted by researchers at Tufts University’s Digital Planet group, which tracked hate speech on Twitter before and after Musk took over. The analysis found that, in the weeks following Musk’s takeover, hateful tweets became much more prominent among the most popular tweets with potentially toxic language: Seven of the top 20 tweets, including keywords that could indicate anti-LGBTQ+ or antisemitic intent and one of the most popular 20 associated with racist language, were judged to be hateful. Prior to Musk’s takeover, just one tweet out of the three top 20 lists was deemed to be hate speech.
Musk shared his “amnesty” decision on Twitter last week after taking a poll on the platform in which over 72% of users who responded voted in favor of reinstating suspended users who “have not broken the law or engaged in egregious spam.” Some 3.1 million users took part in the vote, according to Twitter. The Associated Press has reported that Twitter polls, which Musk also employed in the process of restoring former President Donald Trump’s account, are unscientific and susceptible to the influence of bots.
In India, where Twitter’s role as one of the few remaining forums for relatively free speech is hanging in the balance, Soundararajan said that, prior to Musk’s takeover, there was a consensus at Twitter that a “genocidal process” aimed at Muslims was occurring in the country.
“We have an American company that is operating in a genocidal market without any oversight and that is enabling dangerous speech in an incredibly accelerated way because they’ve basically taken out all of the moderation that was preventing even more copious hate speech,” she said.
In early November, Musk tweeted that he would not make any major decisions about content or reinstating banned accounts before convening a “content moderation council with widely diverse viewpoints.” However, he later backtracked on that pledge, claiming that the content moderation council was a condition of what Musk said was a now broken deal with “a large coalition” of activist groups that threatened to “kill Twitter” by choking off ad revenue.
Now, advocates fear the situation will continue to deteriorate. “Without a comprehensive tool to deal with violence on this platform, this is only going to get worse,” said Jessica González, co-CEO of media advocacy organization Free Press.
Twitter did not respond to TIME’s request for comment.