The danger of generative AI is not some distant algorithmic failure, it is the slow erosion of our ability to tell what’s real, to trust our institutions and to think for ourselves at all.

These are the warnings of Mateusz Łabuz in his new book, The End of Reality, published in collaboration with the Association for the Advancement of Artificial Intelligence (AAAI).

“Everything we see, hear, or read can now be the work of AI. In an era defined by collapsing trust in authorities, governments, media institutions, and increasingly even information itself, synthetic content adds yet another layer to the growing sense of confusion and uncertainty,” he explains.

“Every single day, we are collectively tested on our perception – and collectively fail it. Worse still, we are beginning to stop caring about the result and we passively accept failure.”

Who gets to decide what’s true?

For most of history, Łabuz argues, humans controlled how reality was presented and recorded – even if this was flawed.

“What matters is that for centuries it was humans who decided what would be said, what would be written, and how it would be presented to others.

“That situation began to change radically with the rise of the internet and the emergence of social media, which transformed not only the speed at which information circulated but also the way it was selected and presented,” he explains.

Instead, society began to formulate its beliefs and principles inside streams of content created in real time by millions of users, outside of traditional mechanisms of control but powered instead by AI recommendation systems.

“Over centuries, governments, churches, and media institutions decided which information reached us, acting as so-called gatekeepers – a kind of guardians of the flow of information. They could be wrong. They could abuse their power to spread ideology. But at the very least, they functioned as filters, a form of quality control,” he explains.

Łabuz warns that as algorithms began deciding what we see, what we read, and how we interpret it, what was marketed as personalisation became something else entirely: “Algorithms began assuming the role of gatekeepers themselves, deciding what would appear more frequently and what would disappear unseen into the endless stream of content.”

The fundamental issue with this, he argues, is the incentives driving that system were never aligned with truth.

“Emotion, not information, became the fuel,” he says. “That is why today algorithms feed us emotionally charged, scandalous, and often blatantly false content, skillfully harvesting our attention and pulling us into a machine-built reality to monetize every second of our time.”

He traces how recommendation algorithms now reward cheap, mass-produced content, sexualised imagery, xenophobic clickbait, nonsensical brainrot characters, because it is profitable, not because it is true or valuable. The result, he argues, is: “AI slop need not even tell an outright lie to damage our thinking. It is enough that it convincingly simulates some kind of meaning.”

A new author for reality

Łabuz borrows from philosopher Jean Baudrillard’s concept of hyperreality, a condition in which simulations become more convincing, and more emotionally salient, than reality itself. Generative AI, he argues, has turned this from philosophy into daily experience.

“Synthetic feelings and emotions are treated as genuine because they are reproduced with extraordinary plausibility,” he says. “This hyperrealism of representation fuels a hyperrealism of falsehood and the replacement of truth.”

Łabuz imagines a future archaeologist sifting through what he describes as ‘the digital ruins of the first half of the twenty-first century’ and finding a precise turning point.

“The moment when a world of information created almost exclusively by humans gained a new author. Within little more than a decade, this synthetic world, supervised by algorithms and amplified by billions of everyday interactions between humans and machines, began overshadowing the purely human one.”

He worries about what this means for truth, and warns: “Tools allowing anyone to create their own version of reality make it increasingly difficult to preserve collective agreement about what is actually true.”

This means truth, in Łabuz’ view, can be weaponised. He warns: “Actors with malicious intentions can use them as instruments of political manipulation, humiliation, disinformation, and social destabilization. They can erode values and ideals. Undermine authority. Dismantle trust in institutions and democracy itself. They can also strike at the very core of our privacy, exploiting our likeness for fraud, extortion, or the creation of non-consensual intimate materials.”

The human at the centre

Despite his grave concerns, Łabuz sees a glimmer of hope and it lies in human nature. With this hope, Łabuz hopes it brings humanity’s motivation to fervently maintain qualities such as rigorous scientific study, critical thinking and knowing when to ‘ignore’ technology

He says: “Defending our agency does not require a grand rebellion or an act of heroism. It does not demand a dramatic rejection of technological progress or a retreat into an entirely analogue existence. Such a response would be irrational and, in some respects, immature. Rather, it requires preserving the capacity for independent thought, doubt, and personal responsibility without transferring those responsibilities to algorithms.”