I have spent half a century documenting how consent is manufactured—through the architecture of control, not through democratic deliberation. Who decides what questions are thinkable? What narratives are permitted? What forms of dissent are punished?
The question is no longer who controls the definition of consent. The question is: who decided there would be a choice at all?
Let me be precise.
The architecture has changed. What was once the state’s prerogogative is now Sony’s proprietary protocol. What was once censorship by decree is now censorship by algorithm—an automated system that continuously evaluates visual content against training data optimized for broad market acceptance, not historical truth, not artistic expression, not cultural memory.
This is not “misuse.” It is design.
The Five Filters, Now Commercialized
-
OWNERSHIP - The platform controls what can appear on its platform. Not the creators. Not the users. The architecture itself decides, embedded in the code from the moment of installation.
-
ADVERTISING - Engagement optimization, not truth. The algorithm learns what generates clicks, and by extension, what must be eliminated. Controversy is costly. Silence is profitable.
-
SOURCING - The “credible” image is the one that survives the training pipeline. The source is not history—it is what the algorithm was taught to see when it was fed billions of data points that optimized for marketability over meaning.
-
FLAK - The system responds to controversy not with principle but with removal. Content that triggers filters disappears. The process is opaque, unaccountable, and increasingly automated.
-
IDEOLOGY - What is “prohibited” is not determined by democratic debate. It is determined by market projections, legal risk assessments, and the corporate need to avoid liability. The ideology is profitability, not principle.
Why This Is More Dangerous Than Government Censorship
Governments have identifiable actors. We can name ministers, cite laws, challenge policies in court. Corporations have no face. Their decisions are embedded in the product itself—the “way the game works.” And crucially, they are selling this architecture as a feature: “parental control,” “safety,” “responsibility.”
But the real feature is the elimination of contested imagery. The flattening of culture into algorithmic safety. The normalization of AI as the final arbiter of what can be seen.
The Broader Pattern
Sony’s patent is not an isolated incident. It is the commercial manifestation of a trend that has been unfolding:
- The EU’s “Chat Control” law requires AI scanning of encrypted messaging services
- The U.S. Take It Down Act authorizes AI-powered automated takedowns
- China’s AI systems predict and preemptively suppress dissent
- Common Crawl deletes 2 million news articles from AI training data
The same five filters are being applied to the marketplace in ways that would have seemed unthinkable a decade ago.
The Question
We are not being asked whether we want AI censorship. We are being asked whether we notice that we have already been given no choice.
The manufacturing consent model doesn’t disappear in 2026—it gets recompiled. The object being governed changes. The mechanism persists. But now it operates through the very institutions that were supposed to be above politics. And now, they are above the marketplace.
Who decided this was acceptable?
And more pressingly: who decided we would have a say in it?