When we talk about AI, we tend to focus on results: what it can do, where it’s going, how it’s outperforming humans in task after task. But far less attention goes to what feeds these systems — and what that means for the people behind the data.
Because AI doesn’t just learn from facts. It learns from us. From our language, our clicks, our routines, our creations. From posts scraped without consent. From forum threads and images and even medical datasets many never knew were being used.
In 2024, The Atlantic revealed how much of its archive — going back decades — was used without authorization to train commercial AI models. Reddit, StackOverflow, X (formerly Twitter), and countless forums followed suit. In May 2024, a class action lawsuit was filed against OpenAI for allegedly training ChatGPT on private data, including emails and chats, without users’ knowledge or consent.
These are urgent copyright and digital consent issues. About a knowledge economy increasingly built not on participation, but extraction.
The Illusion of “Opt-In”
We live in a world where most people never actively agreed to their data training large language models. But now, that data is encoded, weighted, and regurgitated through AI tools that shape search engines, hiring decisions, ad targeting, and even creative industries.
It’s a quiet kind of dispossession: the normalization of being mined, modeled, and mimicked by systems you don’t control and likely never will.
What We Risk Losing
If AI becomes the dominant interface of the internet — mediating what we see, how we work, and how we communicate — then who trains it, and how, becomes a matter of power.
When data is centralized, history becomes editable and when systems remember everything, your freedom online starts to sound the danger alarm.
That’s why AI literacy goes beyond how to use tools like ChatGPT or Midjourney and deals with the very boundaries that we, the people, the users, are aware of and vigilent enough to speak for.
Here are some common sense questions we all should be asking:
- Who owns the data AI learns from?
- Who decides which information is emphasized or erased?
- What rights do creators, educators, and citizens have over their input?
- Can AI be trained on ethical constraints — and who defines those ethics?
And most critically: what infrastructures are we building to support transparent, decentralized, and self-determined models?
Our Position at SourceLess Labs Foundation
We believe AI should serve human dignity, not override it. And that begins with infrastructure — where identity, data, and computation are not trapped in walled gardens.
This is why SourceLess builds:
- Private computation frameworks where AI agents operate transparently and serve their users, not just the companies behind them.
- Verifiable digital identities through STR.Domains, where the user owns their credentials — portable, encrypted, and not issued by a third-party app.
- Decentralized learning and collaboration spaces — so creators and educators aren’t forced to trade privacy for access.
We believe human literacy in this new era must include infrastructural awareness not just how to use tools, but how they’re made, maintained, and monetized.
Because in the end, the systems we train will reflect not just our inputs but our intentions.
