Guidelines exist to protect AI workers. Why aren't the world's richest companies following them?
Guidelines exist to protect AI workers. Why aren't the world's richest companies following them?
We are witnessing the ‘Great AI Gold Rush,’ but unlike the traditional gold rushes of the past, the danger isn’t just physical, today it is also psychological. Tech oligarchs that run some of the biggest social media platforms (think Meta and X) are pushing ‘Safety’ as a marketing term with no policy while treating the human training phase as a digital sweatshop. The churn and burn of gig workers is seen in the recent purges of workers and advertisements looking for new workers. The top team members are living a life of luxury while the gig workers are lucky only if they have a steady job so that gig work becomes a supplemental income.
The Core Argument:
• The Ignored Blueprint: Reference the Partnership on AI (PAI) Guidelines. Mention that even though industry standards for “Responsible Data Sourcing” exist, the actual platforms (the “middlemen”) are cutting corners on training and support to keep margins high for the Big Tech clients. These corners are often exhibited in tutorials that do not prepare the worker for the task, glitches in software, and purging staff to replace them with a new crop of workers.
• The Training Vacuum: Companies demand “expert-level” accuracy but provide “zero-level” training. They provide a ‘Black-Box’ where feedback is a mystery and rejections are automated. This leaves the human worker to guess how to keep the machine ‘safe.’
• The Ethical Debt: “Every time an AI model is launched without a robust, supported human training layer, an ethical debt is created. The ‘Oligarchs’ get the profit; the ‘Invisible Architects’ pay the interest in burnout and trauma.
The Call to Action:
In our “Mission Control” area over on Reddit, we are documenting every instance where a platform prioritized “speed to market” over “worker safety.”
If you’ve been asked to perform a high-risk task with zero training, send us your “Redacted Receipt.” * What counts as a receipt? Screenshots of confusing/toxic tutorials, “automated” rejection emails that make no sense, or evidence of platform glitches that resulted in lost pay. Make sure to redact all sensitive information and if you are confused, here’s an example:
We are holding the architects accountable. We want to hear your experience so the community knows they are not alone.