"ChatGPT, Google, Meta and Amazon: how artificial intelligence is really powered", 16 June 2025

... The unseen, untold story of this AI revolution is the massive amount of cheap labour needed to shovel all the data into big tech’s search engines ... That cheap labour includes Joan, Michael and Chychy, who live in Nairobi, Kenya. They call themselves data labellers.

15,000kms from Silicon Valley they’re the workers crucial to AI. They give descriptions, images, labels, notes – absolutely every bit information needed for artificial intelligence to learn and grow.

“AI from my perspective human intelligence. It’s just they’re putting the artificial to sell it to the majority because we’ve been quiet for far too long.” says Joan.

“If you say artificial intelligence, I’ll say you can please correct that and say African intelligence because most of this work has been done here in Africa,” Michael adds.

Kenya is known as the Silicon Savannah, the critical link in the AI supply chain.

... Some of the workers we met had to label and describe explicit content so AI can let the rest of us know, that content is too disturbing to watch or read.

... One company with questions to answers is Appen.

Appen was an early starter in the AI industry, initially focused on voice recognition.

These days its focus is data collection including photos for big tech clients.

Appen’s client list has included Microsoft, Apple, Meta, Google and Amazon.

The workers we met in Kenya say they were asked by Appen to provide photos of children, for a few cents. ...

One company with questions to answers is Appen.

Appen was an early starter in the AI industry, initially focused on voice recognition.

These days its focus is data collection including photos for big tech clients.

Appen’s client list has included Microsoft, Apple, Meta, Google and Amazon.

The workers we met in Kenya say they were asked by Appen to provide photos of children, for a few cents.

... Appen’s full statement

Thank you for the opportunity to respond. At Appen, we take the treatment of our global contributor community seriously and are committed to transparency, fairness, and ongoing improvement. We appreciate the chance to address the claims raised in your inquiry and have provided detailed responses to each question — including specific examples, data points, and explanations of our processes and standards.

To be clear, we do not tolerate unfair treatment of contributors and consistently exceed legal and ethical baselines in the markets where we operate. In Kenya, for example, we have established a minimum base rate of $2.00/hour — well above the statutory minimum wage of $0.73/hour — which is applied consistently across our projects, and we work closely with delivery partners to ensure compensation meets or exceeds that floor.

Spotlight response: Given this work is task-based, Appen was unable to clarity how it guarantees per hour payments

We require contributors to provide informed consent when participating in image-based projects and uphold strong privacy and client confidentiality standards. Due to confidentiality obligations, we are unable to disclose the names of our clients. While project-specific information may at times be limited due to binding agreements with clients, we aim to provide contributors with as much clarity as permitted.

Spotlight response: We asked for further clarification regarding consent i.e is it written, verbal etc but were given no further information

On payments, we understand the frustration that some workers feel regarding minimum withdrawal limits set by certain payment platforms. To clarify, we do not impose these limits ourselves — they are set by the payment providers directly, typically to ensure the withdrawal request can cover any transaction fees associated with the preferred payment method. Our system is designed to work within their frameworks, and we are continuously evaluating alternative providers with better terms, including lower minimum withdrawal thresholds.

One of the key reasons we adopted a wallet-based system was to give Contributors the flexibility to choose their preferred payment provider from the available options. This approach allows for more control and accessibility, depending on what’s available in their region. We also want to note that we do not charge any fees for money transfers or currency conversion. All payments are passed through directly to the chosen provider without any markup or commission from our side.

We remain committed to improving the payment experience and are always looking for ways to make it more convenient and accessible for all Contributors.

In regard to our data storage practices, all data is deleted by Appen after it has been delivered to the client, as the copyright only belongs to the end customer.

Spotlight response: Therefore, it is unknown where images end up or if they can be sold to someone else.

We also want to firmly address the allegation that any contributor was locked out of their account due to engagement with your investigation. Our audit records show consistent logins by the individual in question, and no evidence supports the claim of a lockout. More specifically, our audit logs confirm the contributor logged into his CrowdGen account on the CrowdGen for the first time around 9/13/2024, and subsequently on the following dates: Jan 1, 2025, Feb 10, 2025, Feb 20, 2025, Feb 26, 2025, Mar 18, 2025, Mar 19, 2025, Mar 20, 2025, Mar 21, 2025, Mar 24, 2025, Mar 25, 2025, May 30, 2025 ...