Twitter is now using spam-fighting technology to seek out accounts that might be promoting “terrorist” activity, and is examining other accounts related to those flagged for possible removal, the company announced Friday.
The effort signaled efforts by Twitter to automatically identify tweets supporting “terrorism,” reflecting increased pressure from the U.S. government for social media companies to respond to abuse more proactively. Child pornography has previously been the only abuse that was automatically flagged for human review on social media, using a different kind of technology that sources a database of known images.
Twitter also said Friday it has suspended more than 125,000 accounts for threatening or promoting “terrorist” acts, mainly related to the Islamic State in Iraq and the Levant (ISIL), in the last eight months. Social media outlets have long been a tool for recruitment and radicalization by ISIL and its supporters, who by some reports send tens of thousands of tweets per day.
Rep. Adam Schiff of California, the senior Democrat on the House Intelligence Committee, called Twitter’s announcement “a very positive development.”
In January, the White House made good on President Barack Obama’s promise to reach out to Silicon Valley to tackle the use of social media by violent groups. Those include ISIL, which reportedly inspired attackers who killed 14 in San Bernardino, California, last December.
A post on one of the killers’ Facebook pages that appeared around the time of the attack included a pledge of allegiance to the leader of ISIL. Facebook found the post — which was under an alias — the day after the attack. The company removed the profile from public view and informed law enforcement. But such a proactive effort is fairly uncommon.
The Obama administration recently sent several top officials to San Jose, California, including FBI Director James Comey, Attorney General Loretta Lynch and National Security Agency Director Mike Rogers.
Among the issues they discussed was how to use technology to help speed the identification of “terrorist content,” according to a copy of the White House briefing memo obtained by The Associated Press.
“We recognize that identifying terrorist content that violates terms of service is far more difficult than identifying images of child pornography, but is there a way to use technology to quickly identify terrorist content? For example, are there technologies used for the prevention of spam that could be useful?” the memo said.
Free speech advocates have expressed concerns about these efforts, arguing that it should not be up to tech companies to determine what passes the bar for “terrorist” content. Of particular concern is that different countries have differing lists of designated terrorist groups, and that accusations of terrorism are often subjective and political.
Others argue that wiping pro-ISIL accounts off the Internet deprives intelligence agencies of valuable intelligence and prevents anti-ISIL voices from confronting the group’s ideology head-on.
Nevertheless, in late 2015 Twitter began using “proprietary spam-fighting tools” to find accounts that might be violating their terms of service by promoting “terrorism,” sending them to be reviewed by a team at Twitter. That group also now looks into other accounts similar to those reported to them by other users.
Twitter said it has already had seen results, “including an increase in account suspensions and this type of activity shifting off of Twitter.” But it also noted that there is no “magic algorithm” for identifying terrorist content, which is why even humans reviewing the material are ultimately making judgment calls “based on very limited information and guidance.”
“Like most people around the world, we are horrified by the atrocities perpetrated by extremist groups. We condemn the use of Twitter to promote terrorism,” Twitter said in a statement released Friday. It said it would continue to “engage with authorities and other relevant organizations to find solutions to this critical issue and promote powerful counter-speech narratives.”