The security threats posed by artificial intelligence (AI) are an increasing focus among policymakers. Less discussed are protections for the thousands of human content moderators working behind the scenes to ensure the safety of AI tools and social media—the workers who remove violent imagery, hate speech, and child exploitation from these platforms.

Fasica Berhane Gebrekidan, who worked as a content moderator for Meta's Facebook in Kenya from 2021 to 2023, compiled the stories [PDF] of more than 15 of her former colleagues, many of whom struggled to cope with the disturbing images they witnessed on the job, for the Data Workers Inquiry project at the Distributed Artificial Intelligence Institute (DAIR). An investigation by Time showed that OpenAI paid outsourced workers in Kenya less than $2.00 per hour to screen sexually violent imagery involving minors.

Content moderation, such as data labeling and data cleaning, is just one visible part of a broader ecosystem of data work that is increasingly outsourced to low-income countries. Social media companies employ tens of thousands of moderators (around 15,000 for Meta and around 40,000 for TikTok) through third-party companies to facilitate this job. In Africa, this workforce is now experiencing a mental health crisis—a recent study of 134 moderators from 12 African countries found that more than half of content moderators experienced moderate to severe mental distress. This topic is mostly discussed in news media and computing conferences, but rarely in global health spaces. It is important for the global health community to pay attention. Better workplace conditions, preparation, and support are needed to address the mental health challenges that African content moderators are facing.

In the study, the negative effects were even more pronounced for the moderators who worked for Meta, the parent company of Facebook. Approximately 78% reported psychological distress, a figure much higher than has been reported in some populations experiencing war. Also, former moderators experienced higher psychological distress than current moderators, suggesting that the impact of traumatic experiences does not end when the job does.

The study of content moderators' mental health, published in the proceedings of the 2026 CHI Conference on Human Factors in Computing Systems, was a collaboration between experts with scientific expertise and lived experience. To carry out the research, academics collaborated with individuals who are currently working or previously worked as content moderators to design a survey to collect and analyze data on the psychological well-being of content moderators. The team also included researchers working on the Distributed AI Research Institute's Data Workers Inquiry project and activists.

Lack of Government Regulation, Poor Working Conditions Fuel Mental Disorders

Moderators shared the systemic and structural factors contributing to their experiences of lower mental well-being, including a lack of government policies that require companies to create environments that support mental well-being and ineffective corporate wellness programs. The problem starts from the recruitment stage. Social media companies hire moderators through the business process outsourcing (BPO) industry—third-party companies that have contracts to handle the content moderation process. When BPOs hire moderators in Africa, they tend to not properly inform them of the mental consequences that come with the job. The jobs are usually advertised as "customer support" without detailing the extent of extreme graphic content (i.e., violent, suicidal videos; accidents; sexual content) that could harm their mental health. These moderators are usually recruited in one country and move to another (e.g., a person could be recruited in Nigeria and relocate to Kenya or Ghana) to work. They are invited on visitor visas and face expired work permits or expired visas to stay in the country where they work. These stressors, in addition to working under poor conditions with low pay, high performance expectations, and exposure to frequent graphic content several times a day, likely add to their mental distress.

Furthermore, the corporate wellness programs that are in practice in the workplace rarely serve the content moderators' needs but instead worsen their conditions. The moderators in our study reported a lack of confidentiality during counseling sessions, an absence of mental health counselors with expertise in handling traumatic experiences, and ineffective wellness activities. Moreover, management failed to implement clinical recommendations when provided by counselors and refused to change the practices that negatively impact mental health. So, although companies might indicate that they have wellness programs to help moderators, our research shows that they do not address these employees' mental health needs.

Researchers have conducted similar studies using the same clinical outcome measurement tools in Asia, Europe, and North America. The psychological distress reported by moderators was lower than that noted in our study, which might be partly due to access to better mental health services. These differences across geographies might also reflect the broader health policies in place. Whereas young people in African countries like Ethiopia, Ghana, Kenya, and Nigeria pursue content moderation opportunities because of high unemployment rates, the health protections afforded them are far weaker than those available to workers in Europe or North America. For instance, moderators in the United States can skip graphic content, and moderators in Estonia and the Philippines are properly informed about job implications. By contrast, some African governments see BPOs as a key pillar for economic growth [PDF] and often fail to regulate the companies to ensure the health, treatment, and safety of these essential yet hidden workers. Ideally, policies should be consistent across the continent to ensure that companies do not move to other countries with weaker regulations.

Better Support Needed for African Content Moderators

Researchers and moderators recommend holistic interventions that center moderators' experience instead of a top-down approach that has worked in other regions but might not be effective in the African context. These include regular mental health assessments by qualified clinicians, external to the companies, who have expertise in the local cultural and social environments. Also, companies should clearly describe the work that moderators will be expected to do during recruitment and provide anonymized examples of materials that they will view on a daily basis. Any nondisclosure agreements that moderators are required to sign should allow them to seek psychological support from mental health professionals without fear of retaliation.