AIToday
AI Business & IndustryOpenAI BlogPublished: Nov 19, 2026, 06:00 JST

OpenAI essay rejects 'silent consensus' of ceding control to AI

Send on LINE
OpenAI essay rejects 'silent consensus' of ceding control to AI

3 Key Points

  1. What happened

    In a lecture adapted for OpenAI's blog, the author describes a "silent consensus" — the widespread but rarely stated assumption that humans will inevitably surrender primacy over world affairs to superintelligent machines — and says he rejects it.

  2. Why it matters

    He argues the handover needs no AI rebellion, only countless individually reasonable decisions to delegate; retaining control means holding developers and AI users accountable for harm, with guardrails on autonomous agents.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

The essay is adapted from a lecture the author gave as part of the Vik-Bailey Lecture series at Harvard, and it opens against the backdrop of what some are calling "Black October" — a set of critical internet infrastructure vulnerabilities found by frontier AI systems. The author sets that urgent, reactive work aside to focus on longer-term questions, arguing that without a positive vision of our institutional future, we may drift into a macro outcome we would not choose. He points out that the technology industry is full of positive technological visions — flying cars, autonomous vehicles, robots, nuclear fusion, artificial general intelligence — but lacks a positive vision of how institutions will work.

The author locates the "silent consensus" in a thought experiment: a company run almost entirely by AI systems that outperform human competitors, whose owners face a powerful incentive to take their hands off the wheel, and whose logic, repeated across the global economy, could lead humans to forfeit primacy over global commerce. He stresses that nothing in this scenario requires AI to rebel or owners to wish for disempowerment; each delegation can make sense on its own terms. He considers it a consensus not because everyone agrees it is good, but because the forfeiture of human primacy is widely treated as inevitable or hard to imagine an alternative to.

His counterargument rests on what he calls an unapologetic human chauvinism, but deeper still on the past: human primacy rests not on intellect or a monopoly on understanding but on the civilization humans built, the flame we have tended and now steward. He acknowledges he could be wrong — fully autonomous corporations might outcompete human enterprise, making disempowerment a matter of ordinary capitalist logic, with the agricultural revolution as the closest analog — and that many of his smart and caring colleagues believe that future is likely. Even so, he calls for naming the silent consensus aloud and for building machines compatible with human primacy, with accountability, auditable actions, and permissions users can withdraw.

FAQ
What is the "silent consensus"?
It is the author's name for the mostly unspoken belief that forfeiting human primacy over world affairs to superintelligent machines is inevitable, or at least very hard to imagine an alternative to.
Whose views does the essay represent?
The author's own. The post itself notes it reflects the author's views, not necessarily those of OpenAI or its other employees.
What concrete steps does the author propose?
Building agents with intelligible, limited permissions that can be audited and whose authority users can withdraw, plus accountability for developers and users when harm occurs, and testing whether products leave users more capable.
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleOpenAI grants Apollo Research, METR, Transluce and Goodfire on-site audit access