Amid COVID-19 crisis, tech platforms are making a premature bet on AI, as human moderators shift to moderating less sensitive content while working from home
Today, let's talk about some of the front-line workers at Facebook and Google working on the pandemic: the content moderators who keep the site running day in and day out.
The InterfaceCasey Newton
Context & Ripple Effects
Content moderation at Facebook and Google has been a contractor-heavy, office-bound job — interviews with moderators at YouTube and Facebook in 2019 documented the day-to-day strain of the work. The COVID-19 crisis broke that model overnight: with offices closing, platforms are shifting human moderators to less sensitive content and leaning harder on automation, a bet the reporting flags as premature given AI's known reliability gaps.
Zuckerberg's explanation is the key detail: Facebook is using full-time staff for moderation because it cannot provide mental health support to contractors working remotely — an acknowledgment that the outsourcing structure itself, not just the virus, is the constraint. The story sets up a natural experiment whose results show up months later in the divergence between YouTube's doubled removals and Facebook's steep decline once automation took on more of the load.
First-order effects
Facebook and Google moderators are immediately reassigned to less sensitive content categories while working from home, leaving the most harmful material to automated systems and full-time employees.
Facebook's contractor moderators lose access to on-site mental health support, forcing Zuckerberg to pull full-time staff into moderation duty to cover the gap.
The divergence in removal volumes — YouTube's automation doubling takedowns while Facebook's fell — gives platforms and regulators their first hard data on what AI moderation over- and under-removes at scale.
Third-order effects
If the pattern holds, crisis-driven automation becomes permanent: by 2025, platforms are increasingly replacing human moderators with AI even as moderators themselves say the tech cannot reliably identify harmful content — a structural trade-off between cost and accuracy baked into trust-and-safety operations.
The contractor-based moderation model, built on co-located teams and on-site psychological support, is exposed as incompatible with remote work, pushing platforms toward either insourcing sensitive review or accepting thinner oversight of the most harmful content.
The trend: The pandemic is accelerating the shift of content moderation from contractor workforces to AI and full-time staff, turning a temporary public-health workaround into the industry's default operating model.
I wrote about the dilemma tech companies face with the coronavirus and content moderation. Moderators can't come to work — but they can't work from home, either. And so the platforms are making a premature bet on AI https://www.getrevue.co/... https://twitter.com/...
#BreakingNews. We're teaming up w/ @Whatsapp, @Poynter, @WHO, & @UNICEF on a #Coronavirus Info Hub to spread the word, not the virus. It'll offer #factchecking & services for healthcare pros, educators, nonprofits, local govts, & biz. See in @TechCrunch👇 https://techcrunch.com/..…
A woman called Elizabeth warned friends online about taking painkillers when suffering from #coronavirus. Her warning spread like wildfire on @WhatsApp — but is was fake. How @WhatsApp is now the misinformation frontline https://www.politico.eu/... w/ @JanoschDelcker & @zosiawana…
Must-read report just out from @rankingrights: regulating speech on online platforms will only lead to “a perpetual game of whack-a-mole” - tech companies must be held accountable for how content is amplified and targeted #ItstheBusinessModel https://rankingdigitalrights.org/ ...…
We need an online ecosystem that's optimized for the public interest-NOT the needs of advertisers-now more than ever. Content takedown whack-a-mole won't cut it. From me, @ellerybiddle & the whole @rankingrights team #itsthebusinessmodel https://twitter.com/...
Joint Industry Statement from @Facebook, @google, @LinkedIn, @Microsoft, @reddit, @Twitter and @YouTube https://about.fb.com/... https://twitter.com/...
The technology industry is working closely together in coordination with government healthcare agencies around the world on COVID-19 response efforts. Read our full statement. https://twitter.com/...
Reminder that with Silicon Valley you don't have a choice. You have an illusion of choice. These are separate companies on paper, but they operate as one cohesive unit in reality. We have been working tirelessly for years to provide an actual choice and they demonized us for it. …
I think the biggest next step in combating Coronavirus needs to be shutting down Facebook indefinitely and thus slowing spread of nonsensical disinformation
There's SO much misinformation circulating out there. It doesn't help that this story - that ibuprofen makes Covid-19 worse - has been picked up by major news outlets. Please don't share unevidenced stories. https://twitter.com/...
Facebook is actively taking down coronavirus misinformation unless you're a politician, in which case you're allowed to run misleading ads about coronavirus, unless those ads will cause harm, in which case you're not allowed to run them. https://twitter.com/...