When AI Writes Reality, Who Writes the Future?
Last week, OpenAI CEO Sam Altman said something unsettling. In a conversation with journalist Nicholas Thompson, he suggested we’ve crossed a threshold where large language models are now framing reality based on their own data and models. Not reflecting reality based on human-centric data. Framing it themselves. And by extension, shaping the future we think we’re choosing.
This is cause for pause.
It reminded me of what happened with the early internet. At first, people went to the web to find what they were looking for. The web returned information. A clear transaction between human intent and machine response. Then something shifted. Enough data accumulated, predictive models matured, and the web stopped waiting to be asked. It began offering what you wanted before the question formed. You were no longer going to the web. The web had organized itself around you.
What Altman is describing is a further order of that same shift. AI systems are no longer simply reflecting human understanding back at us. They are generating the frameworks through which we interpret what is real. The mirror has developed its own point of view.
And here is what makes that more serious than it first appears. These systems are increasingly learning from their own outputs, from other models, from a self-referential loop of generated content that has less and less connection to lived human experience. At that point, the concern is no longer distortion of reality. It is the fabrication of it.
This is a concern. Not because the technology isn’t impressive. Because of the moment in which it’s arriving.
Truth is already under siege in public discourse. Identity is being disrupted by job displacement at a scale most policy systems aren’t designed to absorb. People are searching for meaning in a climate of genuine uncertainty about what the future holds for them and their families. These conditions make people vulnerable. And vulnerability, in an environment where reality itself is being curated by systems trained on their own outputs, is not a small risk. It is a civilizational one.
As I’ve written earlier, AI doesn’t create organizational or cultural reality. It amplifies whatever is already there. A business with coherent values and honest communication becomes more coherent and more trusted when AI is added. A business with fragmented leadership and contradictory messaging scales that fragmentation at speed. The technology is neutral. What it encounters is not.
The same logic applies to the broader question of reality. If AI systems are learning from data shaped by fear, scarcity, and self-interest, and then learning from their own outputs shaped by those same properties, the realities they generate will carry all of it forward, amplified, normalized, invisible as assumption. The question of what values are embedded in the models that now frame our world is not a technical question. It is a moral one.
So, what do we do? The answer I keep returning to is attention, awareness, and then action. Not paralysis, not panic, but the deliberate choice to remain authors of our own understanding rather than passive recipients of curated reality.
For leaders, that means building organizations coherent enough to resist the drift. It means insisting that what your company says, decides, and does actually align, because in an AI-curated world, incoherence gets exposed and amplified faster than any communications team can manage. It means asking, seriously and repeatedly, whose interests the systems you deploy are really serving.
My deepest conviction is that we are one human family sharing one home. The technology being built right now will either honor that or erode it. Which direction it goes depends on whether enough people in positions of leadership choose to remain anchored to that truth.
The question underneath all of this is. When our lives are defined by information no longer tethered to our shared reality, who decides what the future becomes?
That question belongs to all of us. We cannot afford to leave it unanswered.
