Someone Has to Watch the Worst of Us So You Don't Have to

Someone Has to Watch the Worst of Us So You Don't Have to

Let's start with a number: 15,000 pieces of content per day. That's a commonly cited workload figure for content moderators at major platforms. Fifteen thousand decisions about whether something is too violent, too sexual, too hateful, too real. Eight hours. Fifteen thousand glimpses into what people are willing to do to each other — and to children — when they think no one's watching.

Someone is watching. That's the whole point. And it is absolutely destroying them.

The People Who Keep the Lights On Down Here

There's a useful fiction that Silicon Valley likes to maintain: that platforms are essentially self-cleaning ovens. That AI handles the bad stuff. That the worst content gets caught by some elegant algorithmic net before it ever touches human eyes.

This is not true. It has never been true. And the companies building these platforms know it.

Behind every major social network — your Metas, your YouTubes, your TikToks — there are thousands of human beings doing what the machines can't: making judgment calls about context, culture, and cruelty. They're called content moderators, trust-and-safety analysts, or, in the sanitized internal language of corporate HR, "community operations specialists." Many of them are contractors, not full employees, which conveniently insulates the platforms from liability. They work in Phoenix. In Austin. In Tampa. In outsourced facilities in the Philippines and Kenya, where American companies pay substantially less for the same psychological exposure.

They watch beheadings before your morning coffee. They review child sexual abuse material and then take a mandated fifteen-minute wellness break. They spend years marinating in the absolute worst documentation of human behavior ever assembled in one place, and then they go home and try to be normal.

Most of them can't talk about any of it. Because they signed an NDA.

The NDA Is Doing a Lot of Work Here

Non-disclosure agreements in the tech industry are so standard they barely register as unusual anymore. You sign one when you get hired. You sign one when you leave. You often sign additional ones if you're involved in anything sensitive — and if your job is watching people get murdered on livestream, "sensitive" barely begins to cover it.

The practical effect of these agreements is that the workers most psychologically damaged by platform content are also the workers least legally able to describe what that damage looks like. Former moderators who've spoken publicly — often after leaving the industry and consulting with lawyers — describe symptoms that map almost perfectly onto PTSD: intrusive imagery, hypervigilance, emotional numbness, relationship breakdown, substance abuse. One former Facebook contractor told a journalist she couldn't watch her own kids play at a park without mentally cataloging dangers. Another described being unable to eat meat for months after a moderation shift.

These aren't edge cases. A 2020 study published in the Journal of the American Medical Association found that content moderators showed significantly elevated rates of anxiety, depression, and post-traumatic stress compared to control groups. The research that does exist keeps pointing in the same direction. The companies respond by citing their wellness programs.

The Wellness Perk as Moral Laundering

Here's the grim comedy at the center of this whole situation: Silicon Valley has spent the last decade absolutely obsessed with employee wellness. Meditation apps. Nap pods. Therapy stipends. Kombucha on tap. The industry that invented the attention economy — the one that deliberately engineered products to be as psychologically captivating as possible — also invented the corporate wellness industrial complex as a kind of offset credit.

For most tech workers, these perks are just perks. Nice to have. Slightly absurd. But for the people working content moderation and trust-and-safety roles, they become something more sinister: evidence that the company knows what it's doing to you and has calculated that a meditation app is cheaper than restructuring the job.

Some facilities offer on-site counselors. Former moderators describe these sessions as largely performative — fifteen minutes to debrief before you go back to the queue. One counselor, they noted, was shared across hundreds of workers. The math doesn't work. It was never designed to work. It was designed to appear in a press release.

Meanwhile, the actual engineers who build the platforms — the people writing the recommendation algorithms, the engagement optimization systems, the autoplay logic — work in an entirely different psychological environment. They're not watching the content. They're just building the pipes it flows through. Their wellness perks are real. Their trauma is largely theoretical.

What "Trust and Safety" Actually Costs

The trust-and-safety function inside major platforms is, in theory, the conscience of the company. These are the teams deciding platform policy: what speech is permitted, what gets removed, how to handle coordinated harassment campaigns, how to respond when a genocide is being organized in a Facebook group in a language the moderation team doesn't speak fluently.

It's high-stakes intellectual work done under enormous pressure with inadequate resources, and it has recently become a political target on top of everything else. Trust-and-safety professionals have watched their teams gutted at Twitter/X, reduced at Meta during rounds of layoffs framed as "efficiency," and publicly attacked by politicians who've decided that content moderation is censorship. The people trying to keep platforms from becoming pure accelerant for real-world violence are now also the people being called enemies of free speech.

The psychological toll of that specific position — responsible for everything, blamed for everything, protected by nothing — is its own particular flavor of burnout. Several former trust-and-safety executives have described leaving the field entirely, not because they didn't care, but because caring was becoming incompatible with remaining functional.

The Thing Nobody Wants to Admit

Here's the uncomfortable truth that sits underneath all of this: the current model only works if someone is willing to absorb the damage. The platforms are profitable because they don't fully internalize the costs of what they host. Those costs get externalized — onto moderators, onto contractors, onto the communities targeted by the content that slips through, onto the mental health systems trying to treat people the industry broke.

There are real policy conversations happening about this. The EU's Digital Services Act includes provisions around moderator working conditions. Some US lawmakers have introduced legislation. Former moderators have won legal settlements — most notably a $52 million settlement between Facebook and content moderators in 2020 — but the structural incentives haven't changed.

As long as it's cheaper to burn through a rotating pool of contract workers than to fundamentally redesign how content flows through these systems, the platforms will keep choosing cheap. As long as NDAs hold, the full scope of the damage stays invisible. As long as the rest of us can scroll without seeing what's underneath the scroll, we don't have to reckon with who paid for that comfort.

Someone is watching the worst of the internet right now. Probably several thousand someones, on multiple continents, staring at screens that most of us would walk away from after thirty seconds.

They signed a form saying they can't tell you what it's like.

But you can probably guess.