Altman warned: dictators with AI are more dangerous than we think

Sam Altman, the head of OpenAI, chose the podium of an Indian artificial intelligence summit to deliver one of the most politically charged speeches in recent times. The man behind ChatGPT and the rapid growth of generative AI addressed the global community with a warning that is difficult to ignore: the most serious threat does not come from the technology itself, but from those who will wield it.

Altman spoke about how the industry and society should remain "humble" in their understanding of superintelligence. The word "humility" sounds unusual coming from the head of a company valued in the hundreds of billions of dollars and methodically moving toward the creation of AGI, artificial general intelligence. But behind this rhetorical restraint lies a quite concrete message: we still do not understand how superintelligence will actually work, what its boundaries will be, and, most critically, how to prevent its use for purposes of suppression.

The central theme of the speech was AI democratization — the idea that access to cutting-edge technologies should be as widely available as possible, rather than concentrated in the hands of a few corporations or states. Altman directly pointed to the danger of a scenario in which totalitarian governments gain access to tools capable of dramatically amplifying surveillance, propaganda, and population control. This is not an abstract threat — today there already exist states actively implementing facial recognition systems, predictive policing, and automated censorship. With the appearance of models like GPT-5 and their analogues, the scale of such control could grow exponentially.

The choice of venue for this statement is far from coincidental. India is the world's largest democracy and simultaneously a country undergoing rapid digital transformation. With a population of over 1.4 billion people and an ambitious AI development program, India is becoming a key battlefield between open and closed approaches to technology. Altman, clearly, seeks to secure the support of democratic states in the Global South, offering them a vision of AI as a tool for expanding citizen capabilities rather than limiting them.

However, Altman's position is not without contradictions. Critics have long pointed to the gap between his public rhetoric about democratization and OpenAI's business practices. The company moved from a nonprofit model to a hybrid structure with a commercial core, attracted billions of dollars from Microsoft, and closed access to the weights of its models. When Altman speaks about democratization, a legitimate question arises: what exactly does he mean by democratization — broad access via paid API or genuine openness, where any researcher can study and modify the model? Companies like Meta with their LLaMA offer a fundamentally different model of openness, and this competition of approaches is only intensifying.

The geopolitical context makes Altman's words particularly weighty. The world has entered a period that analysts increasingly call an "AI arms race." China is developing its own powerful models, including DeepSeek, and building a closed technological ecosystem. Russia is increasing its use of AI in military affairs and internal control. Against this backdrop, the question of who determines the rules of the game in the sphere of superintelligence ceases to be philosophical and becomes a matter of global security. Altman, in essence, proposes a framework in which open access to AI serves as a deterrent — logic familiar from the nuclear age, but with fundamentally different dynamics of proliferation.

The call for humility in the face of a technology we ourselves created sounds like a paradox, but it is precisely in this paradox that the essence of the current moment lies. The AI industry is simultaneously confident in the inevitability of superintelligence and acknowledges that it is unable to predict its behavior. Altman did not propose ready-made solutions in India — he identified a problem whose solution will require an unprecedented level of international cooperation. The question is only whether democratic societies will manage to develop this answer before authoritarian regimes learn to use superintelligence for their own purposes.