OpenAI Starts a Team to Ask Who Ends Up Holding the Power

OpenAI has stood up a research group with an unusually broad brief. The new Strategic Futures team, whose work appears on a fresh blog called AI Futures, is asking how a free society should be reorganised to protect individual rights and agency as increasingly capable AI arrives. The team laid out its thinking in an inaugural post this month.

Its central worry is not a rogue superintelligence or mass unemployment, though it does not dismiss those. The team argues that the largest and hardest long-term risk is the concentration of power. As AI systems get good enough to do work that once required people, the reasoning goes, governments and large institutions could grow less dependent on the humans they govern and employ. When the public's labour stops being essential, so does a quiet source of the public's leverage.

Not decentralisation, balance

The team is careful to say it is not calling for some utopian scattering of all power. It frames the goal as a balance of power: rules, institutions and technical choices that keep any single actor, whether a company or a state, from dominating the economic and political systems that AI will increasingly run through. It points toward laws that strengthen individual autonomy and give smaller organisations room to operate. The planned output is a mix of blog posts, papers, videos and podcasts, with an explicit invitation to argue back.

There is an obvious tension in who is raising the alarm. OpenAI is one of a small handful of firms that has accumulated an enormous amount of exactly the kind of power the essay warns about, and critics will note that a company worth hundreds of billions writing thoughtfully about concentration of power is a little like the house publishing a paper on the dangers of gambling. The ideas can be worth taking seriously and the messenger can be self-interested at the same time. Both things are true here.

A theme the industry keeps circling

The worry itself is not new, and it is striking how often it now comes from inside the industry. Mark Zuckerberg has made the argument for open models partly on the grounds that a few firms should not own AI. Anthropic's Dario Amodei has warned that the public's souring mood toward the technology amounts to a crisis of trust. OpenAI's contribution is to name the endgame plainly and ask what a society should do before it gets there. Whether the company that helped build these systems is the right one to referee them is a question its own blog does not answer.