California bill targeting social media hate speech would hand keys to online extremists
Imagine if we had given the January 6 insurrectionists a map of the U.S. Capitol, marked with the location of every security camera, every security officer, every fire alarm and emergency escape. It would even come with a manual on how police officers are trained to prevent an armed insurrection. That’s the metaphorical equivalent of the digital law under consideration in the California legislature.
Assembly Bill 587, now before the Senate, would require social media companies to share their entire playbook on taking down disinformation, hate speech, predatory content and scams. Just like a security map of the U.S. Capitol, that information is a roadmap to evading detection in the hands of bad actors.
Consider that while you’re on social media scrolling through your friends’ posts, Facebook, Instagram and Twitter are using AI programs to search through the hundreds of millions of daily uploads for content that could hurt users. The programs search for coded phrases like “SWP” (supreme white power) and examine images for pictures that might suggest sex trafficking.
These are the equivalent of security cameras that keep dangerous and illegal content from running down the halls of the Internet. Once you tell criminals or hate groups where they’re located or, in this case, exactly what phrases and images they’re looking for, they become less effective. There are well-documented cases of extremists and hoaxers using coded language or changing the images they use once they know what content social media platforms are scanning for.
AB 587 would also go so far as to share the training manuals for in-person moderators who specialize in removing unsafe content online. When AI programs catch content that might be harmful, these are the security guards who get phoned in. Publishing their training manuals so hate groups, scammers and criminals can read them gives bad actors another tool to skirt their policing efforts.
The law is well-intentioned, and its sponsors want to make the Internet a healthier place. There are racists, predators and con artists online, just like there are in real life, and we should be doing everything in our power to stop them from harming social media users.
That’s why the major social media companies have content policies prohibiting or restricting hate speech, radicalization, disinformation, harassment or foreign political interference. By using advanced AI technology, some companies have gotten pretty good at outsmarting the bad guys and taking down malicious content before it spreads. In a recent report, Facebook announced that its AI model was successful at deleting 95% of all hate content posted to the platform before anyone saw it.
It’s understandable that lawmakers want to know more about where the holes are in social media companies’ internal police work. But we shouldn’t ask Facebook and Twitter to publish their security blueprints in a manner that can be accessed by online extremists and criminals.
We’ve seen legislation like A.B. 587 before, in Republican-controlled legislatures like Texas and South Carolina. In those states, conservative lawmakers proposed similar bills in an effort to help hate groups and insurrectionists evade the rules on social media. That’s why it’s surprising to see the same legislation coming up in California, where Democratic lawmakers have a strong record of combating hate crimes and standing up to the radical right.
Instead of hampering efforts to moderate content online, California lawmakers should be working with social media platforms to prevent online radicalization, predatory content and misinformation. No major online platform wants to become a haven for hate or harassment, and their content moderation practices enable them to stay one step ahead of bad actors.