AI isn’t giving you culture for free – it’s deciding what you see.

That shift matters more than most people realize. As Paolo Barbesino highlights the scale and visibility of AI-curated culture, the next layer is unavoidable: visibility becomes control, and control creates liability. This is not theoretical. It is a structural transition already underway.

For years, platforms like LinkedIn, Facebook, and others, have operated on a simple principle – what gets amplified becomes what people believe is important. Algorithms do not reflect culture neutrally, they rank it, filter it, and push certain narratives forward while others disappear. Your personal feed in whatever social, news, etc. app you use is an ideal example of this. Most users understand how algorithms feed them at a surface level. Few follow it to its conclusion.

AI systems extend this model from distribution into generation. That is the break point. When AI is trained on global inputs – art, writing, history, language – it‘s not just learning from culture, it’s building a compressed representation of it. When generating outputs, it’s no longer retrieving culture as it existed rather, it is reconstructing culture based on patterns, weighting, and optimization goals defined by the system.

This is the point where visibility turns into control because whoever defines the training data, the weighting, and the output constraints is no longer just influencing what is seen. They are influencing what can be produced at all. Entire styles, viewpoints, and historical interpretations can be amplified, softened, or effectively erased through system design choices that most users will never see and that in most examples, will not represent your individual views, morals, or value. This amplification control has already been misused in US elections, documented in multiple investigations. At scale, small biases become structural outcomes.

This is not new behavior. It is the same dynamic that already governs social platforms. The difference is scope and authority. A feed algorithm decides what you scroll past. A generative system can decide what exists in the first place when users ask a question, request an image, or explore a topic. That transition carries consequences.

Ownership is the first.

If a system is trained on global cultural output, who owns the derivative result? The individual creators? The platform? The system operator? Today, the answer is unclear by design. But ambiguity does not remove responsibility. It delays it.

Control is the second.

If cultural output is mediated through systems that optimize for engagement, safety, or policy constraints, then those objectives become embedded in what is produced. Over time, this shapes not just what people see, but what they expect. Culture narrows or shifts without an explicit decision point. The system doesn’t just show culture, it sets the boundaries of what culture can become.

Liability is the third.

If an AI system misrepresents history, distorts cultural narratives, or embeds systemic bias, who is accountable? The model? The operator? The data sources? At small scale, this is dismissed as error. At global scale, it becomes governance.

This is the trade-off most users are not accounting for. The convenience is real, the access is unprecedented, but the cost is not zero. Payment comes in data, in control, and in future leverage over how information, culture, and history are presented and understood. Nothing here is free.

The systems that map culture will, over time, define it. And once that mapping is widely adopted, reversing it becomes difficult, because the reference point itself has changed. This is not a warning. It is a direction of travel. Visibility leads to control. Control leads to consequence.

The only real question is whether we recognize the shift while it is still adjustable, or after it has already been normalized.

Author’s Note: AI tools assisted in drafting, editing, and image prep for this piece. All opinions, arguments, and conclusions are entirely my own.