Kenyan workers who helped prepare data for ChatGPT are urging lawmakers to examine how technology companies outsource content moderation and AI-related work. Their petition puts working conditions and worker protections at the center of a debate over the human labor behind AI safety.
Petition calls for scrutiny of outsourced work
The petition asks lawmakers to investigate the nature and conditions of work, as well as the operations of Big Tech companies that outsource services in Kenya. The workers point to companies such as Sama, which has faced litigation over alleged exploitation, union-busting and illegal mass layoffs of content moderators.
They are also calling for legislation to regulate the outsourcing of harmful and dangerous technology work and protect workers engaged in it. The request raises a practical question: when companies rely on outsourced teams to help make AI systems safer, what safeguards should apply to the people doing that work?
Workers describe exposure to graphic material
The petition follows a Time report on Sama workers who helped make ChatGPT less toxic. According to the report, the job involved reading and labeling graphic text, including descriptions of murder, bestiality and rape.
In late 2021, Sama was contracted by OpenAI to label textual descriptions of sexual abuse, hate speech and violence. The work contributed to a tool built into ChatGPT to detect toxic content. The workers say they faced harmful material without psychosocial support, and that the exposure left them with “severe mental illness.”
Their account points to the human steps involved in building content safeguards. The tool may be part of a product, but the underlying task required people to review and categorize disturbing material. The petition asks lawmakers to consider the conditions under which that labor is carried out.
Companies respond to concerns
OpenAI acknowledged that the work was challenging. The company said it had established and shared ethical and wellness standards with data annotators, but did not provide further details about the specific measures.
OpenAI also described human data annotation as one of several ways it collects feedback and guides models toward safer behavior in the real world. A spokesperson said the efforts of researchers and annotation workers in Kenya and around the world had been valuable to AI system safety.
Sama told TechCrunch that it was open to working with the Kenyan government to ensure baseline protections across companies. It said it welcomes third-party audits and that employees have multiple channels to raise concerns. The company also said it had conducted external and internal evaluations and audits to ensure fair wages and a dignified working environment.
A wider question about AI labor
Sama says it counts 25% of Fortune 50 companies, including Google and Microsoft, among its clients. The San Francisco-based company focuses mainly on computer vision data annotation, curation and validation, and employs more than 3,000 people across its hubs, including in Kenya.
Earlier this year, Sama dropped content moderation services to focus on computer vision data annotation, laying off 260 workers. That shift adds context to the petition: the workers’ concerns sit within an industry where outsourced roles and company priorities can change, while lawmakers are being asked to consider protections for people handling dangerous work.
The petition seeks an investigation and rules for this kind of outsourcing. The responses from OpenAI and Sama describe standards, audits and channels for worker concerns; the workers say stronger protections are needed. The lawmakers’ review could bring those competing accounts of working conditions into focus.