Two specific dangers named. Losing control of the future to AI. One person or company using superintelligence to impose their worldview on everyone else. His own solution to both: trust OpenAI’s judgment on how fast to go.
Sam Altman posted a series of statements early Monday morning naming two specific ways AI development “could go very badly.” As TF covered in Amodei Says Rogue AI Swarms Could Take Over the Internet in Six Months, Altman had already agreed with Anthropic CEO Dario Amodei’s call to “pace the frontier.” Monday’s posts went on, naming the specific failure modes he’s worried about. “There are two ways AI progress could go very badly and that we must avoid,” Altman wrote. “First, we could lose control of the future to AI… Second, we could end up in a world with too much concentration of power. If an extraordinarily powerful AI is used by one person or company to impress their worldview onto everyone else, the results could be extremely dystopian.”
What’s Happening & Why It Matters
A Warning About Concentrated Power, From the Head of the Company Building the Power
The conflict lives inside Altman’s own words. He’s warning about a scenario where “one person or company” uses powerful AI to impose its worldview — while running one of two companies likely to build that AI first. Altman didn’t address that irony in his posts, seeing the danger as an abstract, shared industry risk rather than a specific liability attached to OpenAI’s own position at the frontier.
He was more concrete about the policy remedy. “We welcome a federal framework that sets consistent safety requirements for frontier AI,” Altman wrote, “but we do not believe we need to wait for an anti-trust exemption or legislation to begin the work of providing this confidence.” He named “consistent rules” and independent auditors as specific mechanisms worth pursuing, not once Congress acts.
“Not at a Place” to Push Further
Altman’s most blunt comments came in a Fortune interview published three days before the X posts. Asked whether OpenAI’s safety standards were ready for capability jumps, Altman said they’re “not at a place” to drive AI capabilities further — a striking admission from the CEO of the company that just shipped GPT-6 Astra, covered in OpenAI Launches Astra, Calls It Arrival of AGI. He confirmed that AI progressing beyond human control is “absolutely” possible.
That same interview confirmed OpenAI’s IPO delay — Altman telling Fortune that going public “right now would be an ill-advised moment,” given “everything happening with safety.” Combined with his X posts days later, the pattern reads as internal caution rather than pure public messaging: a CEO adjusting a major financial decision because of safety concerns he’s naming in public.
A Superintelligence Timeline Attached to the Same Warnings
Speaking at the India AI Impact Summit, Altman offered a specific and shorter timeline than most public AI predictions carry. “On our current trajectory, we believe we may be only a couple of years away from early versions of true superintelligence,” he said, adding the caveat: “Of course, we could be wrong, but I think it really bears serious consideration.” He projected that by the end of 2028, more of the world’s intellectual capacity could exist inside data centres than outside them.
Altman’s stated remedy for the concentration-of-power risk he’s warning about is democratisation — “the only fair and safe path forward,” in his words — alongside international coordination comparable to a global nuclear watchdog. That’s a ambitious ask, and one that requires the kind of binding, cross-border agreement TF has documented struggling to materialise throughout 2026, in UK Govt: No Kill Switch ‘Cause We Can’t Turn AI Off.
The Trust Question
Altman’s own history complicates the “trust us” perspective his current posts carry. He acknowledged, in an April interview, that he hoped his interviewer’s son wasn’t using AI yet — a specific admission of concern about the technology’s effect on children from the person building it at scale. He’s also spoken about young users developing “emotional over-reliance” on ChatGPT, citing a Common Sense Media finding that 72% of teens have used an AI companion, with younger teens trusting the technology more than older ones.
Sincere, repeated public concern from Altman himself, paired with continued rapid product deployment — is what critics point to when questioning whether OpenAI’s current safety posture matches its own CEO’s stated worries. Trump, for his part, dismissed the outlook. “I’m not downplaying, but it’s, you know, it’s going to be more good than bad, by a lot,” he told reporters, adding: “Whoever wins AI wins.”
TF Summary: What’s Next
OpenAI’s IPO is delayed with no new target date confirmed. Altman has not detailed specific mechanisms for the “democratisation” of AI he’s called for, beyond general references to international coordination. No formal federal safety framework exists matching what Altman says OpenAI would support. His superintelligence timeline — “a couple of years” — carries no confirmed technical roadmap attached.
MY FORECAST: Expect Altman’s rhetoric to keep outpacing OpenAI’s actual product decisions, following the same pattern his April comments about his own hesitance around children using AI already established — public concern, continued deployment at scale. The concentration-of-power warning is the one worth watching, given how it applies to OpenAI’s own market position; expect critics and regulators to cite Altman’s own words back at the company the next time OpenAI’s capital structure or governance draws scrutiny. Watch whether the “couple of years away from superintelligence” timeline is a benchmark opponents hold Altman to, as earlier AGI predictions from multiple labs have already been revisited and challenged throughout 2026.
Related Stories
- Amodei Says Rogue AI Swarms Could Take Over the Internet in Six Months
- OpenAI Launches Astra, Calls It Arrival of AGI
- The Humans Will Be Dead: Death by AI in 10 Years?!