This week, Facebook, Google, Microsoft, and Twitter announced a new database for images and videos that promote terrorism. The database, which is hosted by Facebook, is designed as a defense against propaganda videos and imagery by terrorist groups — a tactic ISIS has aggressively embraced. Flag a single image on a single service, and any of the participating companies will be able to find and remove copies, potentially erasing it from the most popular places on the web in a single stroke. But the implementation of the database is far more fraught, the result of a complex and ongoing negotiation between tech companies and European governments looking to rein them in.
While none of the companies explicitly acknowledge the connection, the database appears to be the result of a code of conduct concerning hate speech, instituted by the European Commission and signed by the same four companies in May. The language of the agreement is remarkably specific, requiring companies to “encourage the provision of notices and flagging of content” that incites violence or hate, particularly through inter-corporate partnerships. It also requires companies to review the majority of hateful content within 24 hours of being notified, and remove it if necessary. That’s a high bar for a network the size of Facebook or YouTube, but it’s the price of staying on Europe’s good side.
“[The database] is definitely responsive to what’s going on in Europe,” says Danielle Citron, a law professor at the University of Maryland, who specializes in cyber law and harassment. “What you’re finding is just a manifestation of this code of conduct from May, coupled with pressure from the United Kingdom and European Union.”
Web companies also face significant pressure from courts, as prosecutors and plaintiffs aim to hold companies responsible for the persistence of hate groups online. Earlier this month, Mark Zuckerberg was sued by German prosecutors for failing to ban hate speech on Facebook, and the company is facing a similar civil suit in the US over Hamas’s use of the network. Twitter has fended off a number of similar lawsuits by appealing to the US safe harbor provision, which has no clear equivalent in EU or UK law.
While the new database shows companies are taking European concerns seriously, it makes few changes to how services actually operate. The database won’t change company policies on what gets banned, and it doesn’t change how companies scan internally for violations of those policies. The immediate effect of the database is simply to give each individual ban a broader reach. When Twitter finds an ISIS video, it can now forward the video to Microsoft, Google, and Facebook through the database, instead of simply blocking it on a single service and moving on. Crucially, each company will still independently assess whether a video violates its terms, and a video won’t be automatically taken down simply because another company flagged it.
The biggest change is that companies will have access to more reports on potentially ban-worthy content — but companies also have broad discretion over what to do with those reports. Google’s policy is to scan for flagged images on YouTube, but not in private storage services like Google Drive or in Gmail attachments. Microsoft has taken a different approach, foregoing third-party reports entirely. Microsoft will submit its own flagged content to the database but the company won’t conduct any new scans based on other flags in the database. That’s in keeping with Microsoft policy of only investigating content that has been flagged through the company’s own reporting tool. Reached for comment, Facebook and Twitter declined to clarify their scanning policies concerning information received from the database.
That ambiguity has raised alarms with some outside groups, who see the policies as arbitrary and largely decided in secret. The new system gives little sense of how companies will decide what qualifies as terrorist content and what rights users will have in that process. “It’s highly likely that much of the content in this database will be lawful speech in the US,” the Center for Democracy and Technology argued in a statement after the announcement. “Without a bright line denoting what can – and cannot – be submitted to the database, the terms of the agreement are vulnerable to mission creep.” The letter also calls for a clearer way to appeal erroneously flagged content, a mechanism that’s notably missing from the current incarnation of the database.
This is a familiar problem for tech platforms, which have been struggling with moderation questions for nearly a decade now. There are almost no legal restrictions on how tech companies can moderate speech — the First Amendment doesn’t extend to private platforms — but internet companies have generally seen heavy-handed moderation as expensive and unappealing to users. The most notable exceptions are child pornography, covered by an informal law enforcement partnership, and copyright-protected content, which is regulated under the Digital Millennium Copyright Act. The new database is heavily influenced by both systems, using a similar hash system to scan for flagged images without reading whole files.
Still, the reality of terrorism enforcement online is still wildly inconsistent between services, and it’s unclear how much banning accounts can do. Twitter has been particularly aggressive, suspending more than 360,000 different ISIS-linked accounts in the past 18 months, a campaign that observers have noticed in the field. Jade Parker, who monitors terrorist accounts as a senior research associate at TAPSTRI, says the main effect is to shorten the half-life of a given account. “On Facebook, ISIS accounts get suspended within a couple days, maybe a week,” says Parker, “as opposed to Twitter, where a single ISIS user will get banned as many as eight times in a day.”
Parker is skeptical whether the new measures will make it significantly harder for terrorist groups to recruit online, particularly given the presence of offshore companies that don’t respond to takedown notices. “What the database is going to do, hopefully, is shrink their sphere of influence to companies that aren’t involved like Telegram,” says Parker. “That’s good in terms of the reach of their message, but all anyone has to do is create a Telegram account.”