The artificial intelligence industry has grown rapidly over the past few years amid the emergence of generative AI apps led by OpenAI’s ChatGPT.
Since ChatGPT 3.0’s release in November 2022, several companies have introduced their own brands, with search engine giant Google launching Bard, later rebranded to Gemini, Meta Platforms introducing Llama 3, while DeepSeek launched V3.
The industry is expected to continue growing rapidly, reaching a valuation $4.8 trillion, according to a new UN Trade and Development report. However, reaching such heights does not come without challenges.
While challenges relating to the security of data and user privacy are frequently highlighted as a major area that needs addressing, the cost of building high-quality and more powerful AI models continues to be a major factor.
To get to the valuation that several reports are predicting, AI companies will need to continue building smarter agents, capable of driving the next frontier in technology. That means getting the best AI specialists and engineers and securing enormous graphic processing units to power deep learning models.
Both these cost drivers provide little room for maneuverability for AI companies, which means that to cut costs, they have to look elsewhere. It all comes down to training, which explains why most of the companies have been looking to countries in Africa for a more affordable source of labor required for data collection, cleaning, and labeling.
AI training compensation raises ethical issues in Kenya
Over the past few years, AI companies have sought to set up shop in Kenya in search of affordable labor costs. OpenAI was broadly featured in several reports that highlighted complaints from Kenyans who secured jobs with the leading AI company.
While some reports suggested workers were paid a minimum wage of $12.5 per hour, those who complained had a different story to tell.
Kenyan civil rights activist Nerima Wako-Ojiwa told CBS’s 60 Minutes in November that American companies were taking advantage of desperate workers, leading to “a culture of exploitation” with unfair wages and no job security.
“The workforce is so large and desperate that they could pay whatever and have whatever working conditions, and they will have someone who will pick up that job,” Wako-Ojiwa said.
According to some of the workers, Wambalo, Nathan Nkunzimana, and Fasica Berhane Gebrekidan, who worked for SAMA, an American company that outsourced workers for Meta and OpenAI, they claimed to have received $2.00 an hour compared to the advertised $12.50.
“If the big tech companies are going to keep doing this business, they have to do it the right way,” Wambalo said. “It’s not because you realize Kenya’s a third-world country, you say, ‘This job I would normally pay $30 in U.S., but because you are [in] Kenya, $2 is enough for you.’”
How decentralized AI brings fairness to agent training
While AI companies quoted an average hourly wage of about $12.50, workers in Kenya complained of receiving as little as $2:00 an hour. So why the discrepancy? I got in touch with Dr. Max Li, the founder and CEO of decentralised AI company OORT, which is working to address some of the challenges related to AI data collection and training.
According to Li, who recently spoke at the Conviction 2025 Conference in Vietnam, the problem may not lie with the AI companies themselves, but rather the local agencies tasked with hiring data collectors and trainers in various markets.
Dr. Li on stage at Conviction 2025 Conference, Vietnam.
Max Li
Caption: Dr. Li on stage at the Conviction 2025 Conference, Vietnam. Source
His company, OORT, has built OORT DataHub, an AI data collection and training platform that allows anyone to participate in simple tasks and get paid in a “fair and transparent” way.
The company has established a global data contributor network of more than 330,000, said Li, adding that it is “averaging 100,000 daily active users.”
The OORT DataHub compensation matrix distributes OORT tokens based on tasks completed; each task has its own reward, which, according to Li, is based on the complexity of completing it. Users can then swap their OORT tokens for fiat on various exchanges, including Pancakeswap, UniSwap, Gate.io, Bybit, and MEXC.
Essentially, the platform partners with enterprise AI companies seeking agent training data, whereby they submit their data requirements, which are then displayed on the OORT DataHub for contributors to complete as tasks.
One of the complaints from Kenyan workers was that they are paid less compared to AI trainers from developed countries. According to Li, OORT addresses this challenge by making the reward per task transparent to all contributors. Users know what to expect as a reward before completing a task. “With OORT DataHub, it does not matter where you are in the world, Africa, Europe, South America, or Asia,” Says Li. “Our focus is transparency, efficiency, and fairness.”
The platform also addresses one of the biggest complaints raised against AI, user privacy and data security, by making sure OORT DataHub does not collect sensitive user data, with only an email required to open an account.
“All that is required is a smartphone,” he said, adding that there are no pre-requisite qualifications needed, though you may only participate in tasks you can complete, because some are more technical than others.
OORT is targeting to grow its contributor network to 1 million users by the end of the year, with a key focus on Africa and Southeast Asia. And Li maintains his platform is not doing this just for money, but also “for the common good.”
Through decentralization, blockchain technology is being leveraged to introduce new economic models in some of the biggest industries, including gaming and social media. Decentralisation enables fair compensation of users via tokenomics, whilst also giving them the power to own and control how their data is used.
With platforms like OORT, AI seems to be the next industry up from disruption, as agentic AI continues to grow.
OORT is not the only decentralised AI company targeting the African market. Earlier this year, UK-based Boom Technologies launched PHOTON, a pan-African initiative focused on establishing decentralised, gigawatt-scale AI data centres across the continent.
PHOTON’s goal is to “democratise access to AI infrastructure, enabling African businesses, startups, researchers, financial, medical and academic institutions to benefit from ‘African Intelligence’ – AI that embodies Africa’s history, values and languages,” Boom wrote in an announcement in April.
“Photon ushers in Africa’s AI Sovereignty, empowering it with the capacity to design, develop, host and deploy AI systems that reflect and advance the continent’s own narratives, values, and interests,” said Peter Alfred-Adekeye, Founder of Boom Technologies and PHOTON. “We envision an Africa not confined to low-level tasks like training AI models for $2 an hour, but leading the global AI revolution with innovative solutions that are inclusive and deeply rooted in our rich African identity and culture.”
In summary, different AI companies are trying to address the AI user and training challenges as part of an initiative to bring about a fair and transparent compensation model.
Data is a primary variable in AI training and machine learning, and developing countries in Africa and Southeast Asia have the manpower required to power the industry to the next frontier.