In a recent interview, OpenAI CEO Sam Altman made headlines by comparing GPT-5’s development to the Manhattan Project—raising both eyebrows and alarms across the tech world.
During an episode of This Past Weekend with Theo Von, Altman described testing GPT-5 in unusually dramatic terms. Despite being at the helm of its development, he admitted to feeling “nervous” and even “scared” by the model's capabilities.
He went as far as likening the experience to the Manhattan Project, a reference that has historically symbolized irreversible technological breakthroughs with global consequences.
“There are no adults in the room,” Altman warned, pointing to the absence of proper oversight structures as AI continues to advance at breakneck speed.
But his comments have triggered more confusion than clarity.
While he expressed anxiety over the system’s power, Altman stopped short of offering technical specifics.
What exactly spooked him remains unclear.
Instead, listeners were left with vague warnings, unsettling metaphors, and little substance—prompting skepticism about whether the concern was genuine or part of a dramatic marketing narrative.
A Pattern of Alarm and Ambition
This isn’t the first time Altman has signaled discomfort with the trajectory of AI. He’s previously said that AI could “go quite wrong,” and that companies like OpenAI carry a deep responsibility in how they build and deploy such tools.
However, these warnings often come paired with the rollout of increasingly powerful products.
GPT-5, expected to surpass GPT-4 in speed, intelligence, and adaptability, could mark a major step toward what some call artificial general intelligence (AGI).
But Altman’s rhetoric—tinged with both awe and anxiety—raises critical questions about whether even those building AI are equipped to handle what they’re creating.
Power Without a Plan?
Altman’s comments strike at a deeper concern: who decides how AI is used—and what happens when that power grows faster than our ability to manage it?
Even if his statements were exaggerated for effect, they underscore a troubling contradiction. On one hand, OpenAI positions itself as a pioneer of responsible AI.
On the other, its CEO publicly admits to being unsettled by the very tools his team is producing.
In the end, GPT-5 will likely debut with friendlier interfaces, smarter tools, and perhaps a new logo—but the bigger story lies beneath the surface.
What kind of future are we building when those closest to AI’s development are the ones sounding the alarm?
