Altman Warns: AI Could Be Controlled by a Few Powerful Players, Leaving the Public Without a Voice.

In a podcast released on August 23, OpenAI CEO Sam Altman said one of his biggest concerns is that artificial intelligence could eventually be controlled by a small number of companies, models, or individuals, leaving most people with little real influence over how the technology shapes the world.

Altman argued that this is the wrong destination for AI. In his view, the better path is one in which people retain meaningful control over the future, and society evolves alongside AI systems through ongoing interaction, rather than having key decisions dictated by a handful of powerful players.

Altman warns against concentrated control of AI

At the center of Altman’s remarks was a broader concern about concentration. He said AI development should not be dominated by a tiny group of organizations. If technological capability, product design, and social impact are all determined by only a few firms or individuals, ordinary users may be shut out of the process of shaping the rules and norms that will govern the future.

Altman’s core position is that people should have deeper agency over the future, with society and AI models adapting together over time.

That perspective reflects a more participatory vision of AI governance—one in which the public is not merely expected to accept the outcomes of technological change, but has a genuine role in influencing its direction.

The trade-off between freedom and safety

Altman also acknowledged another risk: AI systems themselves could become difficult for humans to control. But he said the response to that danger matters. According to him, if society becomes so focused on preventing loss of control that it is willing to trade away large amounts of freedom in exchange for safety, it may create a different problem altogether.

His concern is that such an approach could push AI power even further into the hands of a few dominant institutions, increasing the likelihood of monopoly-like outcomes. In other words, an aggressive safety-first framework, if poorly designed, could end up narrowing access and concentrating authority rather than protecting the public interest.

A debate that also reflects tensions with Anthropic

Reports linked Altman’s comments to the stance taken by Anthropic CEO Dario Amodei. Amodei has repeatedly warned that advanced models could pose what he described as very real threats to cybersecurity, financial systems, and critical infrastructure, and he has argued for stronger government oversight.

Altman appears to see that line of thinking as part of a broader tendency to sacrifice openness and freedom in pursuit of safety. The disagreement highlights a major divide within the AI industry: how to manage serious risks without allowing regulation or market structure to lock control into the hands of only a few organizations.

The split over open weights

The same divide is visible in the debate over open-weight models. Anthropic was the only major company last month that did not sign a joint letter supporting open weights.

At the same time, reports noted that China has used the open-weight approach to keep pace with Silicon Valley at lower cost. Nvidia, Meta, and Microsoft have also called on the United States to respond.

OpenAI itself has not open-sourced its most advanced frontier models. Even so, the company did release two open-weight models last year: gpt-oss-120b and 20b. That places OpenAI in a somewhat mixed position—protective of its top-tier systems, but still engaged in the broader open-weight conversation.

AI should expand personal autonomy, not just corporate power

Beyond governance and competition, Altman also said that U.S. AI leaders have not done enough to explain how the technology can actually increase people’s autonomy in everyday life. He suggested that this is one of the most important parts of the public conversation, and one that has been underemphasized.

He pointed to what he sees as the beginning of the largest small-business startup boom in history, driven by AI-enabled tools. That view aligns with his larger argument: the real value of AI should not be measured only by how much it amplifies the capabilities of a few large corporations. It should also be judged by whether it gives a much wider group of people the ability to create, make decisions, and build businesses on their own terms.

Taken together, Altman’s comments outline a clear concern about the direction of AI power. For him, the defining question is not just whether AI becomes more capable, but who gets to shape it—and whether the public will remain an active participant in that process.