Washington(CNN) Days after the Pentagon announced it was investigating the leak of more than 50 classified documents that turned up on social media sites, dozens of them remain viewable on Twitter.
Photos of the leaked documents first appeared on a social media platform called Discord weeks earlier and sat largely unnoticed before making their way to more mainstream sites like Twitter and other platforms like Telegram.
The fact that these documents sat undetected for so long and in some cases continue to circulate online underscores the limited authority the US government has to force social media companies to remove content, even classified materials that threaten national security. It also highlights the subjective enforcement of those companies’ policies that determine what content does and doesn’t belong on their sites.
“This all underscores that the notion that platforms are neutral is absolutely bogus,” said Justin Sherman, founder and CEO of Global Cyber Strategies, a DC-based research and advisory firm. “Doing nothing is still a decision, and platforms have to make judgment calls about what they want to allow on their platforms.”
In Twitter’s case, the classified material appears to have fallen through a policy loophole. The posting of classified US military documents would likely not be a violation of Twitter’s hacked materials policy, a former Twitter employee told CNN, “because there isn’t credible evidence establishing the likelihood of a technical hack or intrusion as the source of the materials.”
Twitter’s content moderation policy allows for the platform to label tweets that share potentially doctored materials, but the team responsible for producing such labels has been gutted under Elon Musk, the former employee told CNN, who spoke on the condition of anonymity to avoid retaliation by Musk.
“The fact that Twitter is unresponsive to questions about this at a time when Musk says he’s interested in public trust does not inspire a lot of confidence,” Eddie Perez, a former Twitter executive focused on product safety, told CNN.
Twitter’s personnel issues have only exacerbated its inability to respond to the classified information incident, Sherman said.
“Twitter has been a content moderation dumpster fire since Elon Musk took over and fired or pushed out or caused people to quit in large numbers,” he said, “and there is basically nobody left paying attention to many of these issues.”
Twitter did not respond to a request for comment. Musk broached the issue in a tweet last week, saying sarcastically: “Yeah, you can totally delete things from the Internet — that works perfectly and doesn’t draw attention to whatever you were trying to hide at all.”
Without additional information from outside sources such as government or intelligence officials, tech companies will struggle to proactively respond to classified material, said Katie Harbath, a former Facebook official who helped lead the company’s global election efforts until 2021.
“There’s the hack part, the leak part, and the classified part,” Harbath said. “There’s all those different components. No company alone is going to be able to determine every single one of those without help.”
Disorder on Discord
Like Twitter, Discord does not appear to have a specific policy against posting classified information. It has roughly two dozen other policies in its community guidelines, including bans on “illegal behavior,” conduct that infringes on others’ “intellectual property or other rights,” and spreading false or misleading information.
On Discord, content resides on community-managed chatrooms known as “servers.” The basic structure of the platform can make it easier for illegal content to linger in niche corners compared to more centralized platforms.
That structure also arguably made it easier for the sensitive classified documents to slip under the radar for weeks.
In 2022, the shooter who killed 10 people in a Buffalo, New York, grocery store posted racist screeds and memes to a private Discord server before embarking on his rampage, but Discord did not learn of or remove the content until after the killing spree and once the shooter began sharing links to the private server to his followers.
The classified documents’ murky origins weren’t just an issue for the platforms.
“Most [Discord] users distributed the files because they thought they were fake at first,” one Discord user who followed the initial flurry of postings of classified documents told CNN. “By the time they were confirmed as legitimate, they were already all over Twitter and other platforms,” said the user, who spoke to CNN on the condition of anonymity.
A Discord spokesperson said in an emailed statement that the company is “cooperating with law enforcement” in response to the “to the apparent breach of classified material.”
“When we are made aware of content that violates our policies, our Safety team investigates and takes the appropriate action, including banning users, shutting down servers, and engaging with law enforcement,” the Discord spokesperson said.
Discord has grown more popular in recent years and now claims 150 million active users, many of them avid video gamers.
Austin Berglas, a former senior FBI official, said the FBI is likely looking to obtain information on which Discord users were administering the chat rooms where the documents appeared and the users that posted the documents.
It is unclear whether the US government warned social media platforms to be on alert for posts containing the specific classified information, and it is uncertain how the government could have done so without divulging the fact of the compromise or the contents of the documents to a wide range of otherwise unauthorized individuals.
“In a scenario where you identify and expect a platform to remove leaked classified documents, you’re asking that company to do classified document verification,” said Sherman. “That is not something your average moderator, especially for these companies that farm out their content moderation to people in low-resource countries that they underpay, are going to be able to do.”
Beyond the difficulty of verifying whether the content is authentic, there is also the issue of whether posts have intrinsic public interest value. In the past, some platforms have used that defense to justify hosting content that otherwise might violate their policies, including social media posts by Donald Trump.
The challenges involving US-based social media companies such as Twitter and Discord become even thornier when the platform is based outside of the United States, experts said.
“Good luck getting any of this removed from Telegram,” said Sherman, referring to a social media platform with operations outside of the US.
