Altman: Private Companies Should Not Be More Powerful Than the US Government

Sam Altman picked a Friday evening, the classic slot for news you would rather keep out of broad discussion, for one of the most significant announcements in OpenAI's history. The company has signed a contract with the Pentagon to deploy artificial intelligence technology in classified systems of the US Department of Defense. The scale and details of the agreement have not been disclosed, but the partnership itself, between the largest developer of generative AI and the military arm of a superpower, changes the rules of the game for the entire industry.

To understand why the announcement landed like a bolt from the blue, it helps to recall the history. OpenAI was founded in 2015 as a nonprofit research lab with a mission to ensure that artificial general intelligence benefits all of humanity. Only a few years ago, the company followed a strict policy of refusing to work with military bodies. But in January 2024, OpenAI quietly changed its usage policy, removing the explicit ban on military and intelligence applications. That drew a wave of criticism at the time, but it seemed more like a theoretical shift. Now theory has become practice, in the form of a Pentagon contract.

Aware of the backlash the news would provoke, Altman held a public Q&A session. The central thesis of his remarks was unexpected from the head of a technology corporation: private companies cannot and should not have more power than a democratically elected government. According to Altman, if a country's legitimate authority decides to use certain technologies for national security, a private company has no right to place its own principles above that decision. In effect, Altman formulated a doctrine of subordinating corporate ethics to state sovereignty, a position that would have been considered unthinkable in Silicon Valley five years ago.

This rhetoric deserves close scrutiny. On one hand, it has a rational kernel. Tech giants have indeed accumulated unprecedented power, and the idea that an elected government should keep control over strategic technologies draws support across a broad range of political forces.

On the other hand, Altman's argument conveniently coincides with OpenAI's commercial interests. Military contracts are a stable and generous source of funding. The Pentagon spends hundreds of billions of dollars every year, and even a small share of that budget can transform the finances of any technology company.

For OpenAI, which keeps burning enormous sums on model training and infrastructure, access to defense money is not just a contract but a strategic resource for survival in the race against Google, Anthropic and Meta.

The competitive context matters just as much. In 2018, Google faced an internal employee revolt over Project Maven, a program analyzing drone video for the Pentagon, and was forced to back off. The landscape has changed radically since then. Palantir, Anduril and other defense technology companies are actively integrating AI into military systems, and Google itself returned to work with the defense sector long ago, if less publicly. Altman is, in essence, acknowledging a new reality: in an era of geopolitical rivalry between the US and China, refusing to work with the military is not a moral stance but a competitive loss. The technology will be built anyway; the only question is who builds it.

However, Altman's statement raises troubling questions about limits. If a democratically elected government decides to use AI for mass surveillance or autonomous weapons systems, should a company comply in that case too? Where is the red line? Altman gave no clear answer, limiting himself to general words about the importance of safety and a responsible approach. Critics are already pointing out that the formula "the government decides, we comply" is not an ethical position but the abandonment of one, packaged in democratic rhetoric.

Deploying AI in the Pentagon's classified systems also raises the question of transparency. The public will probably never learn exactly how OpenAI's models are used in defense projects. That creates a paradox: a company building one of the most powerful technologies in history is at the same time becoming less and less accountable to the very people it supposedly works for.

OpenAI's Pentagon contract is a point of no return, not just for one company but for the whole artificial intelligence industry. The era when technology companies could present themselves as neutral toolmakers is definitively over. AI has become a weapon of geopolitics, and every major developer is forced to pick a side. Altman has made his choice and dressed it in the language of democratic legitimacy. It remains to be seen whether society will accept that choice, or at least the employees of OpenAI itself.

*Meta is designated an extremist organization and banned in Russia.