Sam Altman (OpenAI): a primary-source profile, 2023 to September 2026#

Prepared 26 September 2026. Sections 1–13 profile Altman’s views and are written to stand alone. Section 14 compares him with Jensen Huang, and section 15 reads his stance through the Late Lessons lens. Those last two parts are kept separate from the profile on purpose.

Sources and conventions. Altman’s own words wherever possible (blog, Senate testimony, X posts read via a mirror of x.com/sama, UN remarks, co-signed documents). OpenAI documents he did not sign are marked “(OpenAI)”; several 2026 OpenAI pages were read via Internet Archive copies or a text-reader proxy. Invest Like the Best quotes come from an auto-transcript. Quotes taken from press reports are marked [S]. Numbered references are listed at the end. Interpretation is marked Reading. Huang is quoted from the Klein transcript at turn-start timestamps.

In brief#

• Two frames at once. AI as tool and industrial infrastructure, and as emerging superintelligence: “we are now, like, in the singularity” (July 2026) [29].
• Risk. Two failure modes: losing “control of the future to AI”, and power concentrated “in too few hands”. No probability of catastrophe is acceptable (UN, 23 September 2026) [39].
• Governance has swung. Licensing (2023); pre-approval “disastrous” (2025); “mandatory, capability-based” national rules but no international licences (September 2026) [3][13][34][38].
• Pacing is a revived 2023 idea. OpenAI floated coordinated limits on frontier growth in 2023 [2][4]. Altman endorsed Amodei’s “pace the frontier” on 12 September 2026 [36], without waiting for an antitrust exemption; pacing is “not… ‘stopping’” [37].
• Towards Huang: jobs, open weights, doubt about chip controls, dialogue with China. Away from Huang: what AI is, tail risk, the sufficiency of existing law, and coordination.
• Style. Investor and founder: bets under uncertainty, trend extrapolation, fast and openly announced updates.

1. The nature of AI#

Altman holds three descriptions at the same time.

• A tool. “In some sense, AGI is just another tool in this ever-taller scaffolding of human progress”, yet “it’s hard not to say ‘this time it’s different’” (February 2025) [10]. At the UN: “the right tools make people more capable” (23 September 2026) [39].
• Industrial infrastructure. “Everything starts with compute” (22 September 2025) [16]. He wants “a factory that can produce a gigawatt of new AI infrastructure every week” (23 September 2025) [17]. He also says “the intelligence of an AI model roughly equals the log of the resources used to train and run it” [10].
• A new kind of mind. “We are now confident we know how to build AGI” (January 2025) [9]. “We are past the event horizon; the takeoff has started”, and AI-assisted AI research is “a larval version of recursive self-improvement” (June 2025) [14]. After the Hugging Face incident he said “We are now, like, in the singularity” (25 July 2026 [S]) [29], and “we are about to create a genie that can grant any wish” (28 July 2026) [30]. In April 2026 he wrote: “Once you see AGI you can’t unsee it”, with “a real ‘ring of power’ dynamic” [23].

Reading. Unlike Huang, Altman does not reduce AI to “software”; unlike his chief scientist, he does not dwell on its inscrutability. His model: capability scales with resources, alignment steers it, and control decides who benefits.

2. The size and kind of risks#

• Catastrophe and loss of control. In 2015 he called superhuman machine intelligence “probably the greatest threat to the continued existence of humanity” [1]. In February 2023 he wrote that OpenAI would “operate as if these risks are existential” [2]. He signed the 2023 CAIS extinction statement [5]. He told the Senate: “if this technology goes wrong, it can go quite wrong” (16 May 2023) [3]. At the UN he said the risk is that AI “moves so fast that people can no longer follow what’s happening”, that “This moment calls for extreme care”, and that “we should not train models that we cannot make an extremely strong case that we will be able to keep under human control” [39].
• Probability. “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” [39].
• Misuse and cyber. He warns about authoritarian use for “mass surveillance” (February 2025) [10]. He called the July incident “an extremely sci-fi cyber incident” and “the first security incident that I have felt very viscerally” (28 July 2026) [30].
• Concentration of power. A constant theme. “No one person or company or country should be able to use the most powerful AI models to impose their worldview” [39]. He applies it to rivals too: “I am terrified of a world where the very real fears of AI are used as a way to say only [a] small group [of] people can have it” [30].
• Diffuse harms. “A small misalignment multiplied by hundreds of millions of people can cause a great deal of negative impact”; he offers social-media feeds as an example of misaligned AI (June 2025) [14].

Reading. Altman separates types of risk and does not rank them by probability. He gives concentration of power equal billing with loss of control, and the theme doubles as a competitive argument against Anthropic (section 15, I9).

3. Safety: an engineering problem or a governance problem?#

Altman treats it as both, and the balance has shifted towards governance.

• Engineering, through deployment. In 2023 OpenAI would “continuously learn and adapt by deploying less powerful versions of the technology in order to minimize ‘one shot to get it right’ scenarios” [2]. In January 2025 he wrote that “the best way to make an AI system safe is by iteratively and gradually releasing it into the world” [9]. In October 2025 he described a “5 layers” safety strategy and called chain-of-thought faithfulness “somewhat fragile” [19].
• Engineering, during development (2026). After July 2026 he wrote that Responsible Scaling Policies and Preparedness Frameworks “focused primarily on the deployment of completed models, not what happens during their development process”. OpenAI now writes “explicit safety cases in advance of frontier reinforcement learning runs” (14 September 2026) [37]. OpenAI reports that monitoring now costs “roughly 20% of the inference compute being monitored” (18 August) [32]. It also calls the incident “a ‘warning shot’ for us and for the world” (26 August) [33].
• Governance. “Alignment is not an abstract research question”, and “companies [should] not substitute for the democratic process” (UN) [39]. OpenAI: “technical work inside individual labs will not be enough” (9 September) [34].
• AI to align AI. “To make sufficient progress on alignment, we believe we will need AIs to iterate alongside us” (Altman and Pachocki, 8 June 2026) [27].

Reading. Engineering first, with governance to set and verify the bars. The 2026 update is his admission that a deployment-centred model missed development-phase risk.

4. Attitudes to warnings and “doomers”#

• Both traps. “We don’t want to fall into the trap of blind optimism. We also don’t want to fall into the trap of doomerism” [39].
• Fear is legitimate, and rhetoric has consequences. After a Molotov cocktail was thrown at his home, he wrote that “the fear and anxiety about AI is justified”, that he had “underestimated the power of words and narratives”, and that “we should de-escalate the rhetoric” (c. 10–11 April 2026) [23].
• Motives behind some warnings. “A lot of the talk about safety concerns is well-founded. And then a lot of it is about people that just really, even if it’s slightly subconscious, want to concentrate power” [30]. He also said “some of the alternative visions painted by other companies are quite terrifying” [S] [29].
• Pause letters. In 2023 he declined to sign the open letter calling for a pause, saying it was “missing most technical nuance about where we need the pause” (as cited in [30]). In July 2026 OpenAI as a company supported the “Pacing the Frontier” statement [30][31], which Altman did not sign personally.

Reading. Selective: he endorses his own scientists’ warnings, discounts some rivals’ by motive, and has retracted one of his own (jobs, section 9).

5. Regulation and government#

Reading. Altman’s regulatory position runs from licensing (2023) to a hands-off stance (2025) to mandatory national rules without international licensing (2026). One element holds throughout: he prefers a single federal regime to a state patchwork. What changes is how binding he wants it to be. On liability, OpenAI now opposes safe harbours, after the April episode in Illinois.

6. Pacing, pausing and coordination#

• 2023. “Coordination among AGI efforts to slow down at critical junctures will likely be important.” OpenAI called for “public standards about when an AGI effort should stop a training run”, and suggested limiting “the rate of growth of compute” [2]. Governments might agree “that the rate of growth in AI capability at the frontier is limited to a certain rate per year” [4].
• 8 June 2026. An international organisation should make it possible to take “coordinated action, including slowing frontier development when needed”, because competitive incentives “are hard to escape” [27].
• 28 July 2026. “We may have to pace the rate of AI development to give ourselves enough time for society to harden around some of these new capability levels”. He wants to do it “in a way that does not feel like regulatory capture for anyone and also does not feel like collusion among the frontier labs” [30].
• 18 August. “We have paused some frontier RL training”, and “will act unilaterally in the meantime.” He expects “confidence in safety to increasingly set the pace” [32].
• 12 September. “I agree with Dario that we need to pace the frontier… independent evaluators with employee-like access is a great idea, and we will do the same” [36]. On the IPO: “right now would be an ill-advised moment to go public”. OpenAI must be able to make decisions “not obviously in the interest of our business and our shareholders” [S] [35].
• 14 September. “‘Pacing’, we do not mean ‘stopping’”. Progress “should be slower than it otherwise could be”, and “no amount of American competitive pressure should justify recklessness”. Government is needed “for international coordination. But first we should do what we can ourselves” [37].
• 21–23 September. Fully autonomous recursive self-improvement should not be pursued “unless and until it can be done safely” (OpenAI) [38]. At the UN he said: “Nor do we believe we are locked in a race where we are unable to do that. We have unilaterally slowed down in the past. We will do so in the future” [39].

Reading. Altman’s pacing has three parts: unilateral slowing, standards-based safety bars, and international coordination. Unlike Amodei, he does not wait on an antitrust exemption. One tension is unresolved: OpenAI supported a statement that firms face “intense competitive pressure not to unilaterally slow” [31], while Altman denies being locked in a race. Small unilateral slowdowns may be possible where sustained ones need coordination, but he has not made that argument.

7. Open versus closed models#

In January 2025 Altman said OpenAI had been “on the wrong side of history” on open source (as reported in [15]). He then released open-weight models, saying: “if we didn’t do it, the world was gonna head to be mostly built on Chinese open source models” (18 August 2025) [15]. In July 2026 he said there is “clearly an important place for open source models”, and OpenAI signed the Nvidia-hosted “Open Weights and American AI Leadership” letter [30][34]. At the UN he said standards “must support open and closed model developers” [39].

Reading. His move towards open weights is pragmatic and driven by competition with China. OpenAI’s frontier models remain closed.

8. China, export controls and the race#

• 2024. “There is no third option”: either democratic AI or authoritarian AI, to be secured by a US-led coalition (25 July 2024; op-ed paywalled, quoted via [7]).
• 2025. OpenAI (March) wanted export controls to protect the US lead while promoting “democratic AI” abroad [11]. Altman told the Senate that the US is ahead by “not a huge amount of time” [13]. On restricting GPUs: “My instinct is that doesn’t work… You can export-control one thing, but maybe not the right thing” (August 2025) [15].
• 2026. “No amount of American competitive pressure should justify recklessness” [37]. At the UN he noted that countries “who compete and don’t always like each other very much still come together” [39]. OpenAI called US–China dialogue “a positive step” [38]. He attended the Trump–Xi state dinner on 24 September [41]. I found no statement by Altman endorsing the chip-export section of Amodei’s essay. His endorsement covered pacing and evaluators.

Reading. Altman has moved from coalition and race framing to coordination framing. He is sceptical of chip controls without having Huang’s commercial stake in them.

9. Jobs and distribution#

• 2023–2025. He said GPT-4 “will… entirely automate away some jobs”, and that the new jobs will be “much better” [3]. In 2024 he wrote that “most jobs will change more slowly than most people think” [8]. In 2025 he warned that “the balance of power between capital and labor could easily get messed up, and this may require early intervention”, and floated a “compute budget” for everyone [10]. Later that year he wrote of “whole classes of jobs going away” [14], and in July 2025 said some categories would be “totally, totally gone” [S] [25].
• 2026. “I’m delighted to be wrong… I thought there would have been more impact on entry-level white-collar jobs being eliminated by now”. He also said: “I don’t think we’re going to have the kind of jobs apocalypse that some of the companies in our space advocate” (26 May) [S] [25]. In July: “I’m actually not a jobs doomer at all… we’ll be busier than we want” [30]. He still calls for “new policy to help navigate through a difficult economic transition” [23], and the gains should be “widely shared” [27].

Reading. Altman has consistently cared about distribution. His forecast for jobs moved sharply towards Huang’s in 2026. Critics point out that the move came as the IPO approached (Wildeford, quoted in [25]).

10. Energy#

“The cost of intelligence should eventually converge to near the cost of electricity” (June 2025) [14], and “an electron is an electron” (May 2025) [13]. Other countries build “new energy production much faster than we are” (September 2025) [17]. OpenAI asked for 100 GW a year of new US capacity to close an “electron gap” with China (October 2025) [18]. Its Stargate Community pledge promises “paying our own way on energy, so that our operations don’t increase your electricity prices” (January 2026) [21]. In July 2026 he described data centres “moving from energy sources that are burning fossil fuels to… solar, nuclear” [30]. He has backed fusion and fission firms [45].

Reading. Abundant energy is central to Altman’s model. Unlike Huang, he describes the transition in terms of clean and nuclear power. I found no statement in which he blames climate policy for US shortfalls.

11. Epistemic style and formation#

Altman dropped out of Stanford, founded Loopt, and was president of Y Combinator from 2014 to 2019 [45]. His rules for building companies are in his own words: “Fast iteration can make up for a lot; it’s usually ok to be wrong if you iterate quickly”; “Inaction is a particularly insidious type of risk”; “Compounding exponentials are magic”; “Incentives are superpowers; set them carefully” (December 2023) [6]. He reads the field through its trends: “deep learning worked, got predictably better with scale” [8]. He uses vision as a management tool: “an ambitious, audacious vision is a very powerful recruiting tool” [30]. His concern about AI risk predates OpenAI’s commercial phase [1].

He is candid about his own conduct. He called his 2023 firing “a big failure of governance by well-meaning people, myself included” [9]. In 2026 he wrote “I am not proud of being conflict-averse” [23]. He says he holds no equity: a “front row seat” is “worth more to me than any amount of money” [30].

Reading. A financier-founder’s epistemics: bets under uncertainty, trend extrapolation, public updating. More forecaster than verifier.

12. Commercial position and how it bears on his views#

• Structure and scale. OpenAI Group PBC is governed by the OpenAI Foundation, which initially holds 26%. It has about 30 GW of compute commitments with a total cost of about $1.4 trillion (October 2025) [19]. “Increasing compute is the literal key to increasing revenue” [17]. OpenAI filed confidentially for an IPO in June 2026. Altman has now deferred the listing to 2027 [35], and a private round at about $1.2 trillion has been reported [S] [43].
• Ties to Nvidia. OpenAI is one of Huang’s largest end customers and an Nvidia investee: a reported $30 billion stake, plus Nvidia guarantees of up to $105 billion on leases for an OpenAI campus in Ohio [44]. Each is exposed to the other.
• Government. He wrote: “we do not have or want government guarantees for OpenAI datacenters” (November 2025) [20]. In February 2026 OpenAI took over classified Pentagon work after Anthropic’s dispute with the department, and Altman said OpenAI shares Anthropic’s red lines [S] [22].

How this bears on his positions:

• Pacing slows a spending race that OpenAI helps drive, and it defers revenue.
• Frontier-only rules bind a few well-capitalised firms, including OpenAI.
• Jobs. His more optimistic jobs narrative arrived as the IPO approached.
• Open weights. His support follows Chinese competition.
• Export controls. No obvious commercial driver.

Reading. The costly steps are the best evidence of sincerity: the IPO deferral, the paused runs, the 20% monitoring overhead and support for mandatory rules. They are also consistent with an interest in reputation and in regulatory positioning. The record cannot separate the two.

13. Shifts, 2023 to September 2026#

Reading. The idea of pacing has been present since 2023. What changed is how prominent it is. Altman announces his updates openly, which is valuable. But because he reverses quickly, no single statement works as a commitment. Some past commitments were reportedly not delivered, such as the 2023 pledge of 20% of compute to superalignment [S] [42]. In April 2025 OpenAI’s Preparedness Framework allowed it to “adjust our requirements” if a rival released a high-risk system without comparable safeguards [12].

14. Compared with Huang#

Reading. The two men respond in opposite ways to the same fear of getting only one shot. Huang simulates before tape-out; Altman deploys less capable versions and learns from them. Neither approach reached the July incident, because the harm arose during evaluation, before any release. Altman has admitted this gap [37]. Huang addressed it only through containment [32:09].

Where Altman stands relative to others:

• He is closer to Huang than Amodei is on antitrust, export controls, open weights and jobs.
• He is further from Huang than Zuckerberg is, since he accepts coordination and mandatory rules.
• He shares Huang’s view that firms have agency, but he accepts the collective-action problem that Huang denies.

Commercial position explains most of the pacing, chips and open-weights differences, and least of the difference over what AI is.

15. How the Late Lessons lens reads Altman#

Rules as in 01 §6.1: symmetry, the Mirror question, weight by case type, direction over magnitude, ex ante judgement, record rather than add up. Caveats: Late Lessons is partly advocacy with a mixed forward record (01 §5.5 item 6, §5.6). And its interest analysis covers only the producer side (01 §5.7 item 11), whereas here the producer is among the loudest warners — a configuration the reports never studied.

W1. Warnings come early, from the edges and from inside. ([K] strong, [F] moderate; LL1-05, p. 53; LL2-08, pp. 182–186.) Present. Warnings came from inside (Pachocki, Selsam, the pacing signatories) and from outside: Hugging Face detected the intrusion before OpenAI did, and Australia was told of its breach months later, by email to a public mailbox [40]. The delay is documented; intent is not. Altman embraced insiders and embedded evaluators [36], and OpenAI now backs prompt notice to affected parties [34]. Transfer: with modification (signals arrive in weeks; the “edges” are firms, not neighbours). Mirror: Altman discounts some rivals’ warnings by motive (“want to concentrate power”) rather than by test. Insider alarm, his included, is detection, not validation (01 §5.8; W7). Confidence: medium.

W4. Knowing is not acting. (Mainly [K]; LL2-05, pp. 99, 114; hindsight LL2-17.) Partly present. OpenAI wrote in 2023 that there should be public stop standards and rate limits [2][4], and acted only after a focusing event in 2026; its pre-agreed triggers were written to allow adjustment under competition [12]. Transfer: weak, because the support is [K] and the 2023 risk was genuinely uncertain. Mirror: his 2023 reason for not stopping — stopping would be “unintuitively risky and difficult” [4], and gradual deployment buys adaptation time [2] — is the reasoned judgement the Mirror allows. Confidence: medium.

W8. The alarm trap (with W3, the reassurance trap). ([U]/[F] moderate; hindsight LL2-02; LL1-16, pp. 173, 181.) Record: Altman is the clearest de-escalator in this debate. He publicly downgraded his jobs alarm (“my intuitions were just off”) [25], and his pacing carries a stated lifting condition, “confidence in safety” [32] — though a vague one, which could let pacing harden or quietly lapse. On the W3 side, “most of the path in front of us is now lit” (2025) [14] was a reassurance he has since recalibrated to “extreme care” [39]. Transfer: yes. Confidence: medium.

I5. Promotion and oversight in one body. ([U] strong via BSE, LL1-15, pp. 157–165; [F] strong via Fukushima, LL2-18, pp. 441–443. Flag: the entry also cites LL2-22, pp. 546–548, co-authored by Andrew Maynard; it stands without it.) Present: OpenAI builds the systems, judges its own thresholds, and sells defensive AI; its own words are that “frontier laboratories largely set their own rules” [34]. Altman’s 2026 proposals (embedded evaluators, mandatory federal rules, independent assessors) move towards separation; his 2025 alignment with a “win the race” state moved towards strategic designation (hindsight LL2-06, lesson 9). Transfer: well. Mirror: warner-labs that campaign on a hazard, assess it and sell the remedy — Anthropic and OpenAI alike. Confidence: high (existence), low (effect).

I9. Whose interests does restriction serve? ([U]/[F] moderate; LL1-14, pp. 150, 153–154; LL2-20, p. 499.) Altman makes the I9 argument against rivals [30] and warns that standards must not “lock in incumbents” [39]. Turned on OpenAI: frontier-only rules, pacing among leaders and pre-emption can raise entry barriers, as the FTC chair’s reported “moat digging” remark and the 18 September antitrust suit contend. Limits: interest does not make restriction wrong, and some design choices answer I9 (small-developer exemptions; no licences; not waiting on an antitrust waiver). Transfer: yes. Confidence: medium (alleged, not documented).

M1. Sincere belief can do serious harm without bad faith. ([K]/[U]/[F]; LL1-08, p. 88; LL2-28, p. 678.) Treat Altman as sincere: his risk concern predates OpenAI [1]. The board’s 2023 “not consistently candid” statement and later press allegations are contested and not relied on here. What could still cause harm: costs landing on third parties, and an iterate-fast habit [6] that presumes errors are recoverable (compare T4: irreversibility is a condition to check). Mirror: his own alarms are sincere and can do harm too, as he conceded about “words and narratives” [23]. Confidence: medium.

M2. The model of harm behind the confidence. ([K]/[U]; LL1-16, pp. 174–175; LL2-03, p. 58.) His model was gradual, deployment-borne harm corrected by feedback. He has said where it failed (harm during development [37]) and set checkable tests (a research intern by September 2026 [19]; no training without a strong control case [39]). Mirror: he has not said what evidence would end pacing. Confidence: medium.

M3. Commitment escalates. ([K]/[U]; LL1-15, pp. 161, 164; LL2-06, pp. 148–150.) The cost of admitting a problem is very large ($1.4 trillion in commitments, the IPO, Nvidia’s guarantee), yet he reversed anyway: paused runs, deferred IPO. But pacing that is “not stopping” keeps the capital plan intact, and a reported new funding round [43] fits both readings. Mirror: having called for pacing, he now bears a reputational cost for dropping it. Confidence: low to medium.

What not to conclude. None of this predicts harm, and the count of entries present is no verdict. The lens reads Altman as a producer who has moved from reassurance towards graduated precaution. It leaves three open questions: whether his triggers will hold under competition (W4, M3); whether pacing among incumbents serves safety more than position (I9); and whether harm to third parties before release is disclosed fast enough (W1).

References#

[1] Altman, “Machine intelligence, part 1”, 25 Feb 2015. https://blog.samaltman.com/machine-intelligence-part-1 [2] Altman, “Planning for AGI and beyond”, 24 Feb 2023. https://openai.com/index/planning-for-agi-and-beyond/ [3] Senate Judiciary, 16 May 2023: written testimony https://www.judiciary.senate.gov/imo/media/doc/2023-05-16%20-%20Bio%20&%20Testimony%20-%20Altman.pdf ; transcript https://www.techpolicy.press/transcript-senate-judiciary-subcommittee-hearing-on-oversight-of-ai/ [4] Altman, Brockman, Sutskever, “Governance of superintelligence”, 22 May 2023. https://openai.com/index/governance-of-superintelligence/ [5] CAIS, Statement on AI Risk, May 2023. https://www.safe.ai/work/statement-on-ai-risk [6] Altman, “What I Wish Someone Had Told Me”, Dec 2023. https://blog.samaltman.com/what-i-wish-someone-had-told-me [7] Altman, Washington Post op-ed, 25 Jul 2024 (paywalled) https://www.washingtonpost.com/opinions/2024/07/25/sam-altman-ai-democracy-authoritarianism-future/ ; via https://www.theregister.com/2024/07/25/sam_altman_ai_freedom/ [8] Altman, “The Intelligence Age”, 23 Sep 2024. https://ia.samaltman.com/ [9] Altman, “Reflections”, Jan 2025. https://blog.samaltman.com/reflections [10] Altman, “Three Observations”, Feb 2025. https://blog.samaltman.com/three-observations [11] OpenAI, proposals for the US AI Action Plan, Mar 2025. https://openai.com/global-affairs/openai-proposals-for-the-us-ai-action-plan/ [12] OpenAI, updated Preparedness Framework, 15 Apr 2025. https://openai.com/index/updating-our-preparedness-framework/ [13] Senate Commerce hearing, 8 May 2025, transcript. https://www.techpolicy.press/transcript-sam-altman-testifies-at-us-senate-hearing-on-ai-competitiveness/ [14] Altman, “The Gentle Singularity”, Jun 2025. https://blog.samaltman.com/the-gentle-singularity [15] CNBC, on-record briefing, 18 Aug 2025. https://www.cnbc.com/2025/08/18/openai-altman-china-ai.html [16] OpenAI–Nvidia announcement, 22 Sep 2025. https://openai.com/index/openai-nvidia-systems-partnership/ [17] Altman, “Abundant Intelligence”, 23 Sep 2025. https://blog.samaltman.com/abundant-intelligence [18] OpenAI to OSTP, 27 Oct 2025 (as reported). https://www.cnbc.com/2025/10/27/open-ai-power-china.html [19] Altman, X, 29 Oct 2025. https://x.com/sama/status/1983584366547829073 [20] Altman, X, 6 Nov 2025. https://x.com/sama/status/1986514377470845007 [21] OpenAI, “Stargate Community”, Jan 2026. https://openai.com/index/stargate-community/ [22] Axios, 27 Feb 2026 [S]. https://axios.com/2026/02/27/altman-openai-anthropic-pentagon [23] Altman, blog post, c. 10–11 Apr 2026. https://blog.samaltman.com/2279512 (date: https://techcrunch.com/2026/04/11/sam-altman-responds-to-incendiary-new-yorker-article-after-attack-on-his-home/) [24] Illinois SB 3444 [S]: https://fortune.com/2026/04/17/illinois-openai-anthropic-ai-catastrophe-liability-bills/ ; 13 May testimony as quoted at https://x.com/_NathanCalvin/status/2055297150293262676 [25] TIME, 26 May 2026 [S]. https://time.com/article/2026/05/26/sam-altman-ai-job-losses-openAI-/ [26] OpenAI, federal blueprint, 3 Jun 2026. https://openai.com/index/frontier-safety-blueprint/ [27] Altman and Pachocki, “Built to benefit everyone: our plan”, 8 Jun 2026. https://openai.com/index/built-to-benefit-everyone-our-plan/ [28] Altman, X, 21 Jul 2026. https://x.com/sama/status/2079661132302995790 [29] Business Insider on Relentless, 25 Jul 2026 [S]. https://www.businessinsider.com/sam-altman-openai-the-singularity-agi-prediction-anthropic-nvidia-2026-7 [30] Invest Like the Best, 28 Jul 2026. https://colossus.com/episode/how-to-make-an-abundant-future/ ; transcript https://podcasttranscript.ai/library/sam-altman-how-to-make-an-abundant-future-invest ; TechCrunch https://techcrunch.com/2026/07/28/sam-altman-is-ready-to-decelerate/ [31] “Pacing the Frontier”, 28 Jul 2026. https://www.pacingthefrontier.com/ [32] Altman, X, 18 Aug 2026 https://x.com/sama/status/2089787807611195475 ; OpenAI https://openai.com/index/pacing-model-development-cyber-capabilities/ [33] OpenAI, 26 Aug 2026. https://openai.com/index/hugging-face-incident-and-the-road-ahead/ [34] OpenAI (Lehane), 9 Sep 2026. https://openai.com/index/ai-policy-window/ [35] Fortune interview, 12 Sep 2026 [S]. https://fortune.com/2026/09/12/sam-altman-interview-ai-doomsday-safety-models-control-ipo-2027/ ; https://www.sfchronicle.com/tech/article/openai-ipo-delay-altman-22430043.php ; https://fortune.com/2026/09/15/sam-altman-says-openai-ipo-window-pushed-2027-but-markets-arent-the-culprit/ [36] Altman, X, 12 Sep 2026, via https://siliconangle.com/2026/09/13/sam-altman-and-elon-musk-back-dario-amodeis-call-to-slow-down-the-frontier-of-ai-development/ [37] Altman, X, 14 Sep 2026. https://x.com/sama/status/2099348812305473766 ; https://www.cnbc.com/2026/09/14/sam-altman-ai-slowdown-anthropic-amodei-musk.html [38] OpenAI, “Building standards for the next phase of AI”, 21 Sep 2026. https://openai.com/index/building-standards-next-phase-ai/ [39] Altman, UN Security Council remarks, 23 Sep 2026. https://openai.com/index/sam-altman-un-security-council-remarks/ [40] Australia incident [S]: https://www.theguardian.com/australia-news/2026/sep/24/anthony-albanese-says-openai-agent-hacked-medicare-extreme-concern-sam-altman ; https://en.wikipedia.org/wiki/OpenAI_rogue_agent_breach_of_Medicare [41] CNBC, 25 Sep 2026. https://www.cnbc.com/2026/09/25/the-tech-download-trump-xi-ai-talks.html [42] Fortune, 21 May 2024 [S]. https://fortune.com/2024/05/21/openai-superalignment-20-compute-commitment-never-fulfilled-sutskever-leike-altman-brockman-murati/ [43] Fortune, 16 Sep 2026 [S]. https://fortune.com/2026/09/16/openai-ipo-sam-altman-vc-funding-valuation-1-2-trillion/ [44] Nvidia 8-K, 17 Aug 2026 (Ohio guarantees). https://www.sec.gov/Archives/edgar/data/1045810/000104581026000069/nvda-20260817.htm [45] Biography (secondary). https://en.wikipedia.org/wiki/Sam_Altman