Twitter saw a fresh exodus of employees on Thursday as the company hit a deadline set by billionaire owner Elon Musk for remaining staff to commit to being "extremely hardcore" or leave the company.
Departing employees posted on Twitter under the hashtag #LoveWhereYouWorked, announcing it was their last day at the social network. Twitter has been convulsed with chaos since Musk finalized his $44 billion purchase in late October. Many accompanied their posts with a saluting emoji, which has become a symbol inside Twitter of respect for those who are leaving.
On Wednesday, in an email to staff entitled "A Fork in the Road," Musk said Twitter would "need to be extremely hardcore" to succeed. Those who chose to stay should expect long, intense hours of work. Those who left would receive three months' severance pay, he wrote. Employees were required to choose by Thursday afternoon.
The new wave of departures is adding to fears that Twitter is losing critical expertise in everything from how the site and its servers are run to how it keeps user data safe and complies with regulations to how it handles toxic and illegal content.
Earlier on Thursday, a group of Democratic senators sent an open letter to the Federal Trade Commission urging an investigation of Twitter. They said they were concerned the company may be violating the terms of a settlement with the agency stemming from past privacy violations.
Musk "has taken alarming steps that have undermined the integrity and safety of the platform," the senators wrote.
Former worker warns cuts will have consequences
When Musk laid off half of Twitter's employees just days before the midterm elections, Melissa Ingle was left in limbo.
She was a data scientist on Twitter's civic integrity team, monitoring the platform for tweets that might break its rules against misleading election claims. But she was a contractor, not a Twitter employee. When the cuts happened, she didn't even know who was left to sign her time sheet.
"My boss was laid off, and my boss's boss — the head of the department — quit. So I did not know who my boss was. I didn't know what new assignment I had," she said.
With the end of voting quickly approaching, Ingle and her team worked overtime to flag falsehoods and violating tweets. She says she thinks they did a good job, given the circumstances.
"But at the same time, we're not really sure if the work we're doing matters to the new ownership."
On Saturday, she got an answer: she no longer had a job at Twitter.
"I only found out that I was fired [because] I happened to be looking at my phone at about 5:30 [P.M.], and I got a little pop up that said you've been logged out of one or more systems," Ingle recalled.
Rapid changes disrupt Twitter's business
Ingle and others warn, Musk's rapid changes risk compromising Twitter's ability to deal with toxic content and are already disrupting its business, as chaos spilling over onto the platform threatens its advertising revenue.
"There is a short-sighted view of platforms that can look at trust and safety work and integrity work as a bit of a cost center, and as the people that are trying to drag the company down rather than actually trying to help the company grow for the long term," said Jeff Allen, a former data scientist at Facebook who co-founded the Integrity Institute, a group focused on online trust and safety.
At Twitter, like other mainstream social media companies, this work relies heavily on people.
There are the staffers who set policies, workers like Ingle who develop automated systems to analyze the 37.5 million tweets posted every hour, and most crucially, a large group of content moderators who constantly review posts. They are almost entirely contractors.
Many of these workers have now been laid off or resigned. The first round of cuts slashed 15% of Twitter's trust and safety employees, according to Yoel Roth, who led the division. Two days after the election, Roth quit.
The initial round of layoffs also eliminated Twitter's entire curation team of about 150 people. They played an important role adding context and descriptions to news and events trending on the platform, and curating collections of tweets from authoritative sources to help address misleading or false claims.
It's unclear how many of the contractors eliminated last weekend were content moderators. Twitter didn't respond to questions about details of the job cuts.
But losing even a portion of that workforce would be a blow. Ingle said their work is critical to improving the algorithms she wrote, and to understanding things computers can't, like sarcasm and parody.
Automated systems "need constant input and updating and testing and tweaking, just like any other computer script would need...If there's not enough people to update the algorithms, they become more and more porous," she said. "Automation is a lofty goal, and it's a great goal. But we're just not there yet."
"Either you have content moderation, or you don't have it," said Sarah Roberts, an information studies professor at UCLA who briefly worked at Twitter earlier this year. "You don't just kind of have content moderation. Removing child sexual exploitation material is content moderation."
Ingle is also worried about the worldwide implications as big events loom, from the World Cup, which starts on Saturday, to elections around the globe.
"We get hyper fixated in the U.S. on the U.S. elections, but we dealt with the recent Brazilian elections and we dealt with elections all over the world: Japan, India, the E.U., U.K.," she said. "If this global decline in Twitter happens, it's definitely going to affect democracies around the world."
The upheaval since Musk took over is already evident on Twitter.
Musk himself tweeted a conspiracy theory. Hate speech surged in the days after the deal closed. Accounts that repeatedly share false claims are getting more engagement, according to NewsGuard, which rates the reliability of online news sources.
Its analysis found while those accounts tweeted just 6% more in the week after Musk took control, they experienced a 57% increase in likes and retweets over the same period.
"The kinds of content they were spreading had disproportionately more misinformation than they usually had, and that is what leads to engagement," said NewsGuard co-CEO Gordon Crovitz.
Musk's first big product change — letting users buy so-called blue checks, which previously indicated that high-profile users were who they claimed to be — unleashed a flood of accounts impersonating companies, celebrities and politicians.
White nationalists and far-right extremists also signed up for the checks, according to a review of accounts by the Southern Poverty Law Center.
Twitter's trust and safety team had warned about the potential for the feature to be abused ahead of its rollout in an internal document first reported by Platformer and seen by NPR.
That included "impersonation of world leaders, advertisers, brand partners, election officials and other high profile individuals." The document warned that "motivated scammers/bad actors" would likely be willing to pay for the increased visibility offered by the blue checks.
The team recommended ways to mitigate risk, most of which were not adopted, according to notes on the document.
The blue check debacle exacerbated Twitter's business woes, as more advertisers halted spending. Roberts says it's no wonder big brands are wary — and not just of their messages appearing next to toxic tweets.
"They are concerned about being associated with Twitter itself as a brand," she said.