The Human Cost of a Safer Internet

Trust and Safety work is necessary, high-stakes, and often invisible. The people working at Cinder, and in the broader industry, are asked to make careful decisions about some of the worst things that happen online: abuse, exploitation, graphic violence, harassment, fraud, hate, and everything platforms would rather users never see. Though the work protects communities, it creates an occupational burden for everyone working on it.

That's why wellness in trust and safety cannot be an afterthought.

The Industry Has a Wellness Problem

Workplace mental health is already a global issue. The World Health Organization estimates that 15% of working-age adults live with a mental disorder, and that depression and anxiety cost the global economy 12 billion working days and $1 trillion every year in lost productivity.

The added layer of repeated exposure to harmful content compounds the problem.

Earlier this year, The New York Times Opinion published a short documentary following an analyst at Offlimits as he reviews reports of child sexual abuse material. The film does not show illegal imagery, but it makes the emotional reality of the work impossible to ignore: keeping harmful content off the internet still depends on people who have to encounter it first.

"I've spent my career on the frontlines stopping bad actors. Having a comprehensive wellness program is critical for employee wellbeing and long term success," said Noam Dovas, who works on AI Product at Cinder.

A 2024 study of content moderators found a dose-response relationship between exposure to distressing content and symptoms of psychological distress and secondary trauma. The more frequent the exposure, the higher the risk. A 2025 replication study found that more than a quarter of commercial content moderators demonstrated moderate to severe psychological distress, and about a quarter were experiencing low wellbeing.

That should not surprise anyone who has spent time in this field. But it should change how companies operate. And it has changed how Cinder does.

The Best Support Is Proactive, Not Performative

There is a version of workplace wellness that is mostly branding: a benefits page, a meditation stipend, a sentence in the handbook. This is not enough for Trust and Safety teams, and it's not enough for Cinder.

Some of Cinder's work requires employees to engage directly with disturbing material. One example is red teaming: deliberately probing AI systems to find the ways they can fail or be misused, before real users are harmed. In practice, that often means intentionally trying to get a model to generate the very content it's supposed to refuse, so we can help customers build safeguards that hold. It's adversarial by design: the job is to think like a bad actor and surface the worst outputs a system is capable of. That work is essential to making AI safer. It's also, unavoidably, sometimes hard to sit with.

"Red teaming is one of the more challenging parts of building and testing AI models," says Sahaj Kohli, Cinder's Founding AI Engineer. "I make sure to take care of myself by spending time with my dog Rumi and spending time away from a screen."

At Cinder, we recently launched our internal Wellness Hub to start making that support more explicit. This includes free, confidential 24/7 counseling, a monthly gym reimbursement to support physical health, guidance for coping with exposure to harmful content, wellness support in new-hire onboarding, and shared resources for our teams. Cinder also talks about it out loud, with each other, to normalize letting individuals step away, go for a walk, or take the steps they need for themselves. Cinder's cofounders all worked queues at some point at Meta, and they understand how complex the challenges are.

"This is not the whole answer but a beginning," says Cinder's Founding Recruiter Bridget Martin. "We cannot erase the difficulty of this work but we are making support visible and accessible."

Trust and Safety work is harmful through accumulation: repeated exposure, constant context-switching, and the sustained effort of reviewing material designed to shock, manipulate, exploit, or harm.

As the burden of this work has evolved, the benefits and protections around it need to evolve too. At Cinder, that is the standard we are working toward.