Behind the seamless interface of modern Artificial Intelligence lies a hidden, rural workforce of Indian women enduring psychological trauma to keep the internet safe.
- Invisible Labor: Thousands of women in rural India, labeled “ghost workers,” spend up to 800 hours a day moderating violent and graphic content to train AI algorithms for global tech giants.
- The Psychological Toll: Exposure to relentless abuse, pornography, and violence leads to “emotional numbing,” chronic anxiety, and secondary trauma comparable to risks in lethal industries.
- A Lack of Protection: Despite the $250 million market value of India’s data annotation industry, workers remain unprotected by labor laws, silenced by NDAs, and often lack access to mental healthcare.
On a quiet veranda in Jharkhand, India, the domestic sounds of clinking utensils and footsteps provide a stark contrast to the horror unfolding on Monsumi Murmu’s laptop screen. Murmu, a 26-year-old content moderator, is watching a video of a violent assault. It is a scene so disturbing she instinctively speeds up the playback, yet her job—essential to the “magic” of machine learning—requires her to watch it until the end.
Murmu is part of a growing, invisible workforce in rural India that serves as the frontline filter for the world’s most powerful technology companies. These “ghost workers” classify images, videos, and text flagged by automated systems, teaching algorithms how to recognize—and eventually block—human cruelty. But while the AI learns to be “safe,” the humans teaching it are suffering profound psychological decay.
The Dangerous Myth of “Safe” Digital Work
For global tech firms, rural India has become the ideal “data factory.” Approximately 80% of data-annotation workers come from rural or marginalized backgrounds, where lower labor costs and a pool of first-generation graduates provide a cheap, scalable solution for AI training. Women, in particular, are targeted for these roles. Seen as detail-oriented and reliable, they are encouraged to take home-based contracts that are marketed as “safe” and “respectable” alternatives to manual labor or migration.
Experts argue that this work is anything but safe. Sociologist Milagros Miceli, lead of the Data Workers’ Inquiry, classifies content moderation as “dangerous work,” placing it in the same risk category as lethal industrial jobs. The harm is not physical, but cognitive. Studies show that the relentless exposure to graphic content triggers “heightened vigilance,” intrusive thoughts, and a phenomenon Murmu describes as feeling “blank”—a state of emotional numbing that signals deep-seated trauma.
“God’s Work” and the Trap of Ambiguity
The transition from mundane tasks to traumatic exposure often happens without warning. Raina Singh, a graduate from Uttar Pradesh, initially took a data-annotation job to secure a steady income. Her work began with dull tasks like flagging spam emails. Six months later, she was reassigned to an adult entertainment project where she was forced to categorize graphic pornography and child sexual abuse material for hours on end.
When Singh raised concerns, her managers used a chilling justification, telling her, “This is God’s work—you’re keeping children safe.” This narrative of moral duty, combined with the desperate need for employment, creates a “trap of gratitude.” Research by Priyam Vadaliya suggests that because these jobs are seen as rare opportunities for marginalized communities, workers feel a cultural pressure to be grateful, which actively discourages them from speaking out about the mental health toll.
A Systemic Lack of Accountability
The scale of this industry is massive. In 2021, India’s data annotation market was valued at $250 million, with 60% of that revenue flowing from the United States. Yet, of the eight companies investigated for this report, only two provided any form of psychological support. The others claimed the work was “not demanding enough” to require mental healthcare.
Furthermore, India’s labor laws currently offer no legal recognition for psychological harm in the workplace. This leaves workers like Murmu and Singh in a legal and emotional vacuum. Strict Non-Disclosure Agreements (NDAs) further isolate them, preventing them from seeking support from friends or family for fear of legal retaliation or termination. For Murmu, the fear of losing her £260-a-month salary outweighs the fear of the nightmares that haunt her.
Seeking Solace in a Digital Void
As AI continues its rapid global expansion, the reliance on human moderators shows no signs of slowing down. For the women in Jharkhand and Uttar Pradesh, the “quiet” of the offline world has become their only medicine. Murmu spends her free time walking in forests or painting traditional patterns on her walls—small acts of rebellion against the digital violence she is forced to consume.
“I don’t know if it really fixes anything,” Murmu says of her walks under the open sky. “But I feel a little better.” It is a heartbreaking sentiment from the woman responsible for making the digital world a safer place for everyone else, at the cost of her own peace.