‘The Cleaners’: Who scrubs the internet’s darkest horrors?

The Cleaners remain the invisible workforce scrubbing the internet’s most violent and disturbing content for platforms Americans scroll every day. The 2018 documentary of the same name brought their existence into view, yet the scale of outsourced labor and its mental toll has only grown since. New AI tools and ongoing lawsuits make their future uncertain and their present conditions newly relevant.

Behind the Manila desks

Most of the work happens in Philippine contractor offices where reviewers watch flagged videos in rapid succession. Platforms send the worst material through third-party firms that keep moderators on short-term contracts. The arrangement keeps costs low and accountability distant from Silicon Valley headquarters.

One moderator recalled seeing hundreds of beheadings across a single shift. Reviewers must decide delete or ignore in seconds, often after only a few days of training. Quotas can reach 25,000 images or clips daily, leaving little room for breaks or context.

Non-disclosure agreements prevent workers from discussing specific clips or company policies. Many sign clauses that acknowledge possible trauma before they even begin. The secrecy protects platforms while leaving moderators without outside support networks.

Daily volume and edge cases

Algorithms handle obvious violations such as nudity or graphic violence, but nuanced posts still reach human queues. Facebook’s own reports show hundreds of millions of removals per quarter, with moderators handling the gray-area content. The remaining decisions shape what millions of users never see.

Reviewers must interpret political speech, terrorism propaganda, and self-harm imagery under shifting community standards. A single misstep can trigger supervisor review or contract termination. The pressure encourages quick, conservative choices that can suppress legitimate expression.

Pay historically hovers around local minimum wage, roughly one to three dollars an hour in Manila. Workers receive limited mental-health resources despite repeated exposure to traumatic material. The gap between corporate revenue and frontline compensation remains wide.

Psychological fallout documented

Studies and clinician reports link the work to PTSD, anxiety, insomnia, and emotional detachment. Some moderators describe intrusive memories that follow them home. A few cases have ended in reported suicides, prompting internal reviews at contractor firms.

Facebook settled a U.S. class-action suit for $52 million after moderators demonstrated lasting trauma. European centers, including one in Barcelona, have faced collective lawsuits after multiple staff deaths. The pattern suggests the problem is structural rather than isolated.

Platforms now require some applicants to acknowledge mental-health risks in their contracts. The clause shields companies legally but does little to reduce exposure. Workers often lack follow-up care once contracts end.

AI enters the pipeline

Meta and others have rolled out large-language-model tools that pre-screen obvious violations at higher speed. Oversight Board researchers are testing safeguards against bias and hallucination in automated decisions. The shift promises fewer human eyes on the worst material.

Human moderators remain essential for context-heavy posts that algorithms still misread. Layoffs tied to election cycles in Brazil show platforms trimming staff even as regulatory demands increase. The result is a smaller, more stressed workforce handling complex cases.

Transparency reports indicate automated actions now dominate certain harm categories. Yet edge decisions still land with contractors who carry the psychological load. Efficiency gains have not translated into better support for the people left in the loop.

Regulatory pressure builds

The UK Online Safety Act requires detailed enforcement data from platforms, including human-review metrics. EU proposals on age-appropriate design and labeling add new compliance layers. U.S. antitrust focus has so far sidestepped content-moderation labor issues.

Contractor firms resist unionization efforts that could raise wages or mandate counseling. Platforms maintain that third-party status keeps them at arm’s length from employment claims. Regulators have yet to close that gap.

Advocacy groups push for mandated trauma screening and limits on daily graphic exposure. So far, voluntary guidelines remain the norm. The absence of binding standards leaves workers dependent on corporate goodwill.

Media spotlight since 2018

The Cleaners documentary premiered on PBS Independent Lens and sparked coverage in the LA Times and NPR. Reviewers quoted in the film described the work in religious terms, comparing it to moral cleansing. The framing resonated with U.S. audiences encountering the issue for the first time.

Subsequent reporting tracked the $52 million settlement and European lawsuits. Coverage often highlights the contrast between polished platform safety pages and the outsourced reality. Public awareness has grown, yet hiring practices have changed little.

Tech outlets now track AI moderation pilots alongside ongoing contractor conditions. The story has shifted from revelation to incremental updates on an entrenched system. Reader interest remains steady whenever new legal filings surface.

Cultural ripple effects

Users rarely consider that every removed post passed through someone’s screen. The invisibility reinforces the sense that harmful content simply disappears. In reality, the labor transfers distress from one population to another.

Comparisons to gig-economy burnout appear in online forums, though moderators lack even the flexibility of drivers or delivery workers. The fixed quotas and secrecy distinguish this workforce from more visible precarious jobs. Still, the mental-health conversation overlaps.

Some former moderators have begun writing anonymously about their experiences. Their accounts circulate on niche forums and occasionally reach mainstream outlets. The trickle of testimony keeps the human cost visible despite NDAs.

What platforms claim versus practice

Company statements emphasize safety teams and proactive detection. Internal documents reveal reliance on overseas contractors for volume and deniability. The gap between messaging and operations persists across multiple platforms.

Former policy executives have noted that pre-screening everything would require resources companies have declined to allocate. Flagged-only review keeps costs down but leaves harmful content online longer. The trade-off favors speed and scale over thoroughness.

Shareholder pressure for margin growth discourages investment in higher contractor wages or on-site counseling. Until liability or regulation shifts the cost-benefit calculation, the status quo holds. Public statements alone have not altered hiring or support practices.

Paths still open

Some advocates propose licensing requirements for content-moderation contractors, similar to security firms. Others call for platform liability when contractors fail to provide adequate care. Either route would require legislative action that has yet to materialize.

AI advances may reduce raw exposure, but complex speech and cultural context will still need human judgment. The question becomes whether platforms will retain experienced reviewers or continue cycling through short-term staff. Retention and training investment remain optional rather than required.

Without structural change, the same workers will continue absorbing the internet’s darkest material. Their labor keeps feeds appearing cleaner, yet the cost stays hidden from users and quarterly reports alike.

Forward from here

The Cleaners occupy a permanent blind spot in platform accountability debates. Their conditions reflect choices about cost, speed, and liability that predate current AI experiments. Until those incentives shift, the work and its consequences will remain outsourced and largely unseen.