Sam Altman: What the Record Shows
Overview
Zvi Mowshowitz describes Sam Altman as the CEO of OpenAI.1 Wikipedia describes Altman as "an American entrepreneur and investor" who has led OpenAI since 2019.2
Who should govern powerful AI
In remarks at the United Nations Security Council on September 23, 2026, Altman said key decisions "cannot be made by labs in San Francisco alone."3 Altman says no person, company or country should use the most powerful models "to impose their worldview on everyone else."4
Of the 15 people quoted on whether powerful AI must be governed this way, 8 agree.
AI risk, safety and pace
Altman agrees that AI poses a serious risk of human extinction or permanent disempowerment.5 Altman said it doesn't matter whether people put the risk of catastrophe at 10%, or 1%, or 12%, or .1%, and that none of these levels are "remotely acceptable," offering reasoning only.5 As building AI becomes more automated, Altman says, the pace of progress "could accelerate rapidly."6
On September 14, 2026, Altman agreed with Dario Amodei on the need "to pace the frontier," giving no reasons.7 Altman also backs independent evaluators with employee-like access and pledges to do the same, again with no reasons given.8 Of the 22 people quoted on extinction risk, 14 agree and 6 disagree.
Cyber threats and security
Altman agrees that AI is creating a serious wave of cyber threats that will force major changes to internet security.9 On October 27, 2026, Altman said, "I no longer think we’re going to be able to just patch all the bugs."9
Altman also says AI models "might be able to find" bugs beyond what a hacker is thought able to do, giving no reasons.10 Of the 15 people quoted on this, 14 agree.
Computer memory and the cloud
On October 30, 2026, Altman called memory in personal computers used five percent of the time "quite terrible" from a global view.11 In a supply constrained environment, Altman says, "you want maximum utilization in the cloud."12
Altman explains that memory makers hold back on expanding because they fear "the bubble might burst."13 These memory views rest on reasoning only.111213
Said differently
Altman is quoted agreeing on September 23, 2026, and disagreeing on October 6, 2026, that labs should not ship frontier AI models unless proven safe. At the Security Council, Altman said "we should not train models" without an extremely strong case for keeping them under human control.14 For this, Altman cited a specific case: "We have unilaterally slowed down in the past."14
On October 6, 2026, Altman also said "the world should accept some bad things happening" for the technology's benefits, giving no reasons.15
Sources
- Zvi Mowshowitz :
Sam Altman (CEO OpenAI)
- Wikipedia :
Samuel Harris Altman (born April 22, 1985) is an American entrepreneur and investor who has been the chief executive officer (CEO) of the artificial intelligence company OpenAI since 2019.
the most important decisions cannot be made by labs in San Francisco alone. They must be shaped through democratic processes
(agree on Democratic governance of AI)No one person or company or country should be able to use the most powerful AI models to impose their worldview on everyone else.
(agree on Democratic governance of AI)It doesn’t matter whether people put the risk of catastrophe at 10%, or 1%, or 12%, or .1%. None of these levels are remotely acceptable.
(agree on AI existential risk)As the process of building AI becomes more automated, the pace of AI progress could accelerate rapidly. This moment calls for extreme care.
(agree on AI accelerating AI and science)I agree with Dario that we need to pace the frontier.
(agree on Slowing AI development)Committing to having independent evaluators with employee-like access is a great idea, and we will do the same.
(agree on Independent AI evaluators)I no longer think we’re going to be able to just patch all the bugs. I think we’re going to have to completely switch the defense model.
(agree on AI-enabled cyber threats)AI models might be able to find “bugs that are outside of the normal threat model of what we think a hacker can do,”
(agree on AI-enabled cyber threats)from a global allocation perspective, actually, it's quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer and being used five percent of the time.
(side unclear on local inference)- Quote not reprinted: this page already repeats as much of this article as we allow. (side unclear on local inference)
memory makers aren't willing to expand much because they're afraid that the bubble might burst and they'd have overbuilt
(side unclear on AI market bubble)we should not train models that we cannot make an extremely strong case that we will be able to keep under human control
(agree on Safety standard for shipping AI)“the world should accept some bad things happening” for the technology’s benefits
(disagree on Safety standard for shipping AI)