Sam Altman posted on X on the evening of November 3, 2026 that “People should always be in control of the future.” The post quotes a phrase about AI systems deciding what to reward in humans, and does not say whose phrase it is.
It is Dario Amodei’s, word for word but one, from his 2024 essay Machines of Loving Grace. Here is the full post, the passage it answers, what Altman said on the same theme in October and a day earlier in Abu Dhabi, and what nobody has said yet.
On this page · 6 sectionsOpen
Sam Altman posted on X at 01:40 UTC on November 4, 2026 (8:40 p.m. ET, November 3) that people should always be in control of the future.
The phrase he quotes, about what AI systems think makes sense to reward in humans, comes from Dario Amodei’s October 2024 essay Machines of Loving Grace. Altman does not name him.
In the essay it was one of several possible economies after powerful AI, listed beside a universal basic income and hedged as having tons of possible problems.
Altman adds that refusing to cede control to superintelligence should ultimately constrain AI progress, a line he also took at ADIPEC on November 2.
As of 02:20 UTC on November 4 we found no reply from Amodei or Anthropic. Why Altman raised the line now is not stated.
§ 01What Altman posted
The post went up at 01:40 UTC on November 4 (8:40 p.m. ET on November 3) and carried three sentences. The first: “People should always be in control of the future.” The second says OpenAI is against the idea that people could be “relegated to a role of doing things” that AI systems think make sense to reward in humans, with that last phrase in quotation marks. The third: “People should not cede control of the future to superintelligence, and that requirement should ultimately constrain AI progress.”
By X’s own count it had 162,388 views 38 minutes after it went up. It names no person and no company other than OpenAI.
§ 02Where the quoted words come from
The quoted phrase sits in the Work and meaning section of Amodei’s essay, which he announced on X on October 11, 2024. He is asking what an economy looks like once AI can do most work, and lists possibilities. The first is simple: “It could be as simple as a large universal basic income for everyone.”
The next is the one Altman quotes. Amodei writes that it could be a capitalist economy of AI systems that give out resources to humans “based on some secondary economy of what the AI systems think makes sense to reward in humans”, with a hedge in brackets: “based on some judgment ultimately derived from human values.”
He does not endorse it. Two sentences later he writes: “All of these solutions have tons of possible problems.” Altman’s version drops one word: Amodei wrote “the AI systems”, Altman quotes “AI systems”.
§ 03Altman on human control, dated
The November 3 post is the third time in a month Altman has taken this line in public.
| Date | Venue | What he said |
|---|---|---|
| October 4, 2026 | POLITICO Decoded interview | Accept bounded risks we understand, not the really catastrophic risk, including serious loss of control |
| November 2, 2026 | ADIPEC, Abu Dhabi, with The National | There should be no handoff to AI; AI must serve people, not the other way around |
| November 4, 2026, 01:40 UTC | X post | People should not cede control of the future to superintelligence |
At ADIPEC, Altman told The National’s editor-in-chief that on human control “we think there must be no room for debate”, and that there “should not be a handoff to AI”. In the POLITICO interview, per the transcript Snopes published, he said: “Accept bounded risks, accept risks that we understand in exchange for the benefits and liberty, agency all those things I was just talking about, but don’t accept, like, the really catastrophic risk.”
That October quote has been circulating again. A November 1 post paraphrasing it as the world needing to accept that bad things will happen passed 9.1 million views on X. Snopes rated the attribution correct on October 8; the paraphrase drops his line about catastrophic risk.
§ 04What nobody has said yet
Altman did not say why he raised a two-year-old passage now, and we found no statement that ties it to anything Amodei or Anthropic said this week. As of 02:20 UTC on November 4 we found no reply from Amodei or from Anthropic.
Two things are on the record either way. Amodei offered the AI-rewards-humans economy as one uncertain option among several, not a plan. And Altman has now written down, as OpenAI’s position, that keeping people in control “should ultimately constrain AI progress”, which commits OpenAI to slowing down if the two ever conflict.
§ 05The same question, at desk size
For a team that already hands work to AI agents, this argument shows up at a much smaller scale every day: which actions an agent takes on its own, and which wait for a person to say yes. The answer that holds up in practice is the one both CEOs circle: people set the limits, and the agent works inside them.
§ 06Sources
- Sam Altman on X, November 4, 2026, 01:40 UTC.
- Dario Amodei, Machines of Loving Grace, October 2024, section Work and meaning.
- The National, ‘No room for debate’ over human control of AI, says Sam Altman, November 2, 2026.
- Snopes, Sam Altman said world needs to accept ‘some bad things happening’, October 8, 2026, with the interview transcript.
- POLITICO, Sam Altman Decoded interview, October 4, 2026.
Q1What exactly did Sam Altman say on November 3, 2026?
Three sentences on X: people should always be in control of the future; OpenAI is against people being relegated to doing what AI systems think makes sense to reward in humans; and people should not cede control to superintelligence, which should ultimately constrain AI progress.
Q2Where does the phrase 'reward in humans' come from?
From the Work and meaning section of Dario Amodei’s essay Machines of Loving Grace, published in October 2024. Altman’s version drops one word, the article before AI systems.
Q3Did Amodei propose that AI should decide what humans earn?
He listed it as one possibility among several, next to a universal basic income, and wrote that all of these solutions have many possible problems and would need lots of iteration and experimentation.
Q4Is this a new position for Altman?
No. At ADIPEC in Abu Dhabi on November 2 he said there should be no handoff to AI, and in an October POLITICO interview he said to accept bounded risks but not catastrophic ones, including serious loss of control.
Q5Did Altman really say the world must accept that bad things will happen?
Close, but not in those words. In the October 4 POLITICO interview he said to accept bounded risks we understand, not the really catastrophic risk. Snopes rated the attribution correct on October 8.