Read time: 3 minutes

Exclusive: OpenAI responds to Kenya’s ‘toxic work’ petition

By , Africa editor
Kenya , 18 Jul 2023
Foxglove director Martha Dark.
Foxglove director Martha Dark.

Artificial intelligence company OpenAI has acknowledged to ITWeb Africa that during ChatGPT's safety training, young Kenyans working as content moderators were placed in a "challenging work" setting.

This comes as some of the Kenyans who helped to train the generative AI solution approached Parliament last week, claiming they were exposed to toxic content and now experience post-traumatic stress disorder (PTSD).

The Kenyan content moderators who helped to train the ChatGPT service petitioned Parliament to investigate OpenAI and its local partner in content moderation, Samasource (Sama), a San Francisco-based firm that hires people in Kenya.

The Kenyans have now claimed exploitation, saying insufficient wages were paid for helping to train the chatbot.

The content moderators now want the National Assembly to investigate the nature, working conditions and business practices of Sama. They have also asked that the working practices of other similar businesses operating in Kenya, to which major tech companies outsource their content moderation and other AI work, also be more closely examined.

In response to the claims, an OpenAI spokesperson tells ITWeb Africa: “We recognise this is challenging work for our researchers and annotation workers in Kenya and around the world – their efforts to ensure the safety of AI systems have been immensely valuable.”

The company didn’t offer a solution to the challenge, however. The company spokesperson adds: “Our mission is to build safe and beneficial AGI (artificial general intelligence), and human data annotation is one of the many streams of our work to collect human feedback and guide the models toward safer behaviour in the real world. We believe this work needs to be done humanely and willingly, which is why we establish and share our own ethical and wellness standards for our data annotators.”

The matter has drawn the attention of UK-based Foxglove – a non-profit that fights to “make tech fair”. Foxglove is backing Kenyans who have approached Parliament with the petition, along with the nascent content moderators’ union, which the ChatGPT petitioners are also involved with.

Foxglove is a group of lawyers, technologists and communicators, who claim they want to fight for a future in which technology is used to benefit all people.

Foxglove director Martha Dark comments: “ChatGPT is world-famous as a symbol of AI’s potential. But like Facebook before it, its success is built on an army of hidden and underpaid people who do the gruesome work of sifting through toxic content to make the service safe.

“Moderators in Kenya are forming the first content moderators’ union on the continent to fight back.

“This parliamentary petition is the latest demonstration of the power of organised tech workers. Foxglove supports this movement – and hopes Kenyan MPs will make urgent reforms to the outsourcing model that allows companies like Sama to enable exploitation by foreign tech giants.”

From Kenya, Mercy Mutemi, managing partner of Nzili & Sumbi Advocates, tells ITWeb Africa: “Kenyans have had enough of being big tech’s cash cow, where huge profits are extracted then sent overseas, leaving the young African workers, who made them, jobless and broken.

“I urge lawmakers to listen to these brave former ChatGPT data labellers and immediately investigate working conditions inside Kenya’s content moderation offices.”

In the petition, the former workers of Sama say training the ChatGPT model involved reading and viewing material that depicted sexual and graphic violence, and categorising it accordingly, so that the AI could learn it for safety purposes in its future interactions with people.

“Throughout the contract of training ChatGPT, we were not afforded psychosocial support. Due to the exposure to this kind of work, training ChatGPT, we have developed severe mental illnesses, including PTSD, paranoia, depression, anxiety, insomnia, sexual dysfunction, to mention a few,” the petition reads.

OpenAI did not respond to these specific claims.

The petition further reads: “While Kenyan youths are encouraged to take up jobs in the digital world, the Constitution guarantees protection from this kind of exploitation. The laws of employment and occupational health should be aligned with the Constitution.”

Daily newsletter