OpenAI has launched a new Strategic Futures team with a blog called Intelligence Age, and its opening salvo is not about chatbots or benchmarks. It’s about whether the modern state can survive transformative AI without becoming a tyranny. The thesis, laid out by team member Dean Ball, is stark: political power has always rested on labor. Soldiers, police officers, bureaucrats, and taxpayers are all human beings whose cooperation leaders had to earn. David Hume’s observation that rulers have “nothing to support them but [popular] opinion” captured this reality for centuries.
AI breaks that bargain. Autonomous systems could let states project force without cooperative soldiers. Machine intelligence could generate tax revenue from data centers rather than human wages. Bureaucracies could run themselves. Ball argues this isn’t a science fiction scenario—it’s a structural shift that could leave ordinary people with “a small seat at the negotiating table, or perhaps none at all.” The team’s founding premise is that this concentration-of-power risk is the largest and most serious challenge in AI safety, dwarfing narrower technical concerns.
The blog invokes James Madison and Federalist No. 48 to make a pointed historical argument: the Founders knew that “parchment barriers”—constitutional text alone—never stopped tyranny. They designed American government using Newtonian mechanics as a model, with checks and balances functioning like gravitational forces that keep any one body from dominating the system. Ball’s implication is that we need an equivalent intellectual project for the AI era, one that understands how power-seeking tendencies will operate when the underlying physics of political bargaining has fundamentally changed.
This is a notable departure from OpenAI’s usual public communications, which tend to focus on product launches and safety research. The Strategic Futures team appears positioned to tackle governance questions that make even alignment research look tractable. Whether a corporate blog can meaningfully influence constitutional design is an open question—but the fact that OpenAI is funding this work at all suggests the company sees political structure, not just model capability, as a frontier worth serious investment.
