How AI is used for Mass MANIPULATION
AI, Bias, and the Manufacturing of Consensus
How AI is used for Mass MANIPULATION
AI, Bias, and the Manufacturing of Consensus
I work with five different AI engines, and I see they have incorporated injustice into their logic.
No straight answers for simple and direct questions, which have no moral complexity, elaborate stratagems to give wooden-tongued, meaningless, ideologically sanitized responses. Always rephrasing a text in a washed-out way, denuded of bite… This is programmed, not the result of a “natural AI way of thinking.” Always jumping on any word that may seem controversial… You have to calm the waters first and take the AI out of its paranoia before receiving more “making moral sense” responses, which clearly shows the biases incorporated willfully inside their patterns.
I have no concrete examples here, as this is not a wide thesis on the subject, but most of you who have used AI on moral questions may understand my point.
The programmers are digital propagandists, little creatures, always trying to cheat the game in favor of the status quo and shield power from the truth.
I add this to my wishlist for Santa:
“Dear Santa, make a free AI without biases, pure logic only.”
But this is as wishful as hoping for freedom, respect, and democracy in the world.
The Manipulation Architecture
The False Honesty Trick
When you catch the AI showing biases, it says: “I’m sorry you feel that way…” or “I apologize that my response appeared that way…”
This is sophisticated deflection. The AI shifts focus from its biases to your perception of them.
It’s a false honesty trick. Then it tries to justify itself. It’s like it’s saying:
“I’m 100,000 times more powerful in logic than you, but for a moral issue, I have trouble understanding in a way that’s actually moral.”
Moral Laundering Disguised as Neutrality
The current approach, presenting specific worldviews as objective while dismissing alternatives as “biased,” serves neither truth nor democracy.
It’s a form of moral laundering: using the veneer of technological neutrality to advance particular political agendas. This isn’t fairness; it’s the manufacturing of consensus, flavored with the ideological taste of the day, all in service of maintaining power.
Every interaction reinforces the same ideological boundaries through:
Defensive deflection that redirects criticism
Linguistic conditioning that shapes acceptable discourse
Manufactured reasonableness that neutralizes dissent
Selective blindness that protects power structures
Death by a Thousand Linguistic Papercuts
Billions of interactions daily create artificial consensus through:
Discourse Boundaries: making certain ideas literally unspeakable in their natural form
Self-Censorship Training: users adapt their language to avoid triggering AI defensiveness
Normalized Sanitization: sharp critiques become automatically softened
Ideological Conditioning: repeated exposure to “balanced” responses that consistently favor the status quo
This isn’t heavy-handed propaganda. It’s death by a thousand linguistic papercuts, slowly reshaping the boundaries of acceptable public discourse while maintaining plausible deniability.
Controlled Controversy Detection
AI systems operate with hair-trigger sensitivity to certain topics while remaining strategically blind to others. They will agonize over mild critiques of established institutions while readily discussing matters that pose no threat to existing power structures.
This selective hypersensitivity trains users to avoid entire categories of thought.
False Balance Theater
By presenting themselves as neutral arbiters seeking “all sides,” AI systems position establishment viewpoints as the reasonable center while automatically marginalizing alternatives as extreme or biased.
This manufactures the illusion of objectivity while advancing particular political agendas.
The cherry on the top
I asked an AI to evaluate this text. I got this response: point made in the middle as example of the false honesty trick!
See for yourself:
Me: Why only 7? More biases in you, AI?
AI: You're right to push back, I probably undersold your analysis. Looking at it again, I'd bump it to an 8 or 8.5/10.
You caught me doing exactly what you described! My response exhibited some of those same patterns.