Sam Altman: checked claims

Scores by kind of claim

Each kind of claim is scored on its own, out of 100: the share that held up, with the range it likely falls in. Fewer resolved claims give a wider range.

Factual claims

Statements about the present or the past, checked against the record: did they hold up?

83 77 88

High confidence, 77 factual claims resolved. From 18 documents, 2014 to 2026; 19 documents sampled from 35 candidates found.

Not scored: 28 unresolvable, 18 checkable only against their own records.

See the factual claims
Predictions

Statements about events outside the person's control, checked once they fell due: did they come true?

69 49 87

Low confidence, 5 predictions resolved. From 4 documents, 2017 to 2025; 19 documents sampled from 35 candidates found.

Not scored: 94 unresolvable. 9 more fall due later.
Does not yet allow for how likely each prediction was when made.

See the predictions
Promises

Commitments to the person's or organisation's own future action, with something to deliver: was it delivered?

69 49 87

Low confidence, 5 promises resolved. From 4 documents, 2017 to 2024; 19 documents sampled from 35 candidates found.

Not scored: 1 unresolvable, 1 checkable only against their own records.

See the promises

All types combined: 81 (likely 76 to 86, high confidence), from 87 claims: 77 factual claims, 5 predictions, 5 promises. It mixes kinds of claim in different proportions for different people, so compare people kind by kind.

Sam Altman's claims

238 claims · page 1 of 3

238 claims, sorted by newest said. Choose a claim to read it in full.
Verdict	Kind	Said	Claim	Topic
Unresolvable	Prediction	8 Oct 2026	With lighter AI regulation, people will do orders of magnitude more good than bad with AI. Prediction · 8 Oct 2026	AI policy and governance
Unresolvable	Prediction	4 Sep 2026	AI compute will keep becoming vastly more efficient, with better chips and smarter algorithms, though it will still require huge investment. Prediction · 4 Sep 2026	AI compute, hardware and data
Own records only	Factual	4 Sep 2026	At the beginning of 2020 the world's top token user was an OpenAI employee using roughly 100,000 tokens a month. Factual · 4 Sep 2026	AI compute, hardware and data
Mostly true	Factual	4 Sep 2026	Altman said OpenAI was announced at the very end of 2015 and that its first day of work was around January 2016. Factual · 4 Sep 2026	Their own work or organisation
Own records only	Factual	4 Sep 2026	By 2015 Sam Altman had become convinced that powerful AI was really going to happen and decided to try building it. Factual · 4 Sep 2026	Their own work or organisation
Mostly true	Factual	4 Sep 2026	OpenAI trained GPT-3 in 2020, which Altman considers the first model above a threshold of usability. Factual · 4 Sep 2026	AI capabilities
Unresolvable	Factual	4 Sep 2026	Work that was expected to take a startup 3 months is now probably doable in about 17 minutes with Codex. Factual · 4 Sep 2026	AI capabilities
Not yet due	Prediction	4 Sep 2026	In a decade, people will not talk about AI as much as they do now, because it will be expected everywhere. Prediction · 4 Sep 2026	AI, the economy and work
Unresolvable	Prediction	4 Sep 2026	Altman said he thinks some things are going to go very wrong with cybersecurity unless people act quite urgently. Prediction · 4 Sep 2026	AI safety and alignment
Unresolvable	Prediction	4 Sep 2026	Even with large efficiency gains, much more AI infrastructure will need to be built to avoid AI being a highly priced commodity. Prediction · 4 Sep 2026	AI compute, hardware and data
Mostly true	Factual	4 Sep 2026	A chart exists showing the historical cost of an hour of light at night, with early data points at about 5 hours of labor per hour of light. Factual · 4 Sep 2026	Technology
Not yet due	Prediction	4 Sep 2026	In another 6.5 years, average per capita token use will rise to roughly 100 billion tokens a month. Prediction · 4 Sep 2026	AI compute, hardware and data
Unresolvable	Prediction	4 Sep 2026	Altman said he thinks this, speaking while describing OpenAI's new model and what people can now do with AI, will be the greatest boom in entrepreneurship and small business creation the world has ever seen. Prediction · 4 Sep 2026	AI, the economy and work
Own records only	Factual	4 Sep 2026	Altman said different people at OpenAI would give different answers, but that the rough consensus, to the degree there is one, was that it came together around the time of GPT-4, which he believed was around March 2023 while saying he might be wrong about the timeline. Factual · 4 Sep 2026	AI capabilities
Own records only	Factual	4 Sep 2026	The top individual token usage rose by a multi-million-fold factor between early 2020 and mid-2026, about 6.5 years. Factual · 4 Sep 2026	AI compute, hardware and data
Not yet due	Prediction	4 Sep 2026	In another 6.5 years, the world's top individual user will use about 100 quadrillion tokens a month. Prediction · 4 Sep 2026	AI compute, hardware and data
Own records only	Factual	4 Sep 2026	In mid-2026 the top token user Altman knows of was using hundreds of billions of tokens a month. Factual · 4 Sep 2026	AI compute, hardware and data
Own records only	Factual	4 Sep 2026	The average person on Earth went from using effectively zero tokens to about 100,000 tokens a month between early 2020 and mid-2026. Factual · 4 Sep 2026	AI compute, hardware and data
Resolved true	Factual	4 Sep 2026	Altman said that a system "we now have" is capable of discovering new knowledge, doing science, adding huge economic value and creating entire pieces of complex software. Factual · 4 Sep 2026	AI capabilities
Own records only	Factual	4 Sep 2026	OpenAI knew early on that using more compute made its AI systems smarter and more capable. Factual · 4 Sep 2026	AI compute, hardware and data
Resolved true	Factual	4 Sep 2026	Sam Altman was a sci-fi and AI enthusiast as a child. Factual · 4 Sep 2026	Their own work or organisation
Resolved true	Factual	4 Sep 2026	It is very cheap today to run a 5-watt bulb for an hour. Factual · 4 Sep 2026	Economics and markets
Unresolvable	Prediction	11 Sep 2025	Sam Altman said it currently looks as though AI will not result in a radical concentration of power, though the trajectory could shift and would require adapting. Prediction · 11 Sep 2025	AI policy and governance
Resolved true	Factual	11 Sep 2025	Elon Musk helped start OpenAI. Factual · 11 Sep 2025	Their own work or organisation
Resolved true	Factual	11 Sep 2025	AI-enabled impersonation scams are no longer a merely theoretical concern. Factual · 11 Sep 2025	Other global risks
Own records only	Promise	11 Sep 2025	Sam Altman will double check and follow up with Tucker Carlson to confirm user data is never given to anyone else except under subpoena. Promise · 11 Sep 2025	Their own work or organisation
Mixed	Factual	11 Sep 2025	Current AI models now hallucinate only a little bit. Factual · 11 Sep 2025	AI capabilities
Unresolvable	Prediction	11 Sep 2025	The world is rapidly heading toward people understanding that they need ways to verify calls and images aren't scams. Prediction · 11 Sep 2025	Other global risks
Unresolvable	Prediction	11 Sep 2025	The changes brought by AI will not be nearly as abrupt as the earlier sudden change Altman described as a point in favor of societal resilience. Prediction · 11 Sep 2025	AI, the economy and work
Unresolvable	Prediction	11 Sep 2025	The problem of AI hallucination will be fully solved. Prediction · 11 Sep 2025	AI capabilities
Unresolvable	Factual	11 Sep 2025	Earlier AI models hallucinated all the time. Factual · 11 Sep 2025	AI capabilities
Mostly true	Factual	11 Sep 2025	Elon Musk later decided OpenAI was not on a trajectory to succeed, told its founders it had a zero percent chance of success, and stopped participating. Factual · 11 Sep 2025	Their own work or organisation
Unresolvable	Prediction	11 Sep 2025	Sam Altman said he feels optimistic that the government can be brought to understand the importance of AI privilege and enact it. Prediction · 11 Sep 2025	AI policy and governance
Unresolvable	Prediction	11 Sep 2025	Many current phone and computer customer support workers will lose their jobs to AI, which will do the work better. Prediction · 11 Sep 2025	AI, the economy and work
Unresolvable	Prediction	11 Sep 2025	Advanced AI will not feel like what the interviewer described; it will feel like a really smart computer that may advise people, and people will listen to it. Prediction · 11 Sep 2025	AI capabilities
Unresolvable	Prediction	11 Sep 2025	AI-driven job change will be a punctuated equilibrium moment, with much of it happening in a short period, though not dramatically different from the historical rate over the long run. Prediction · 11 Sep 2025	AI, the economy and work
Mixed	Factual	11 Sep 2025	AI models such as ChatGPT do not do anything unless a user asks them to. Factual · 11 Sep 2025	AI capabilities
Resolved true	Factual	11 Sep 2025	As of September 2025, US law gives no legal privilege protecting users' conversations with AI comparable to doctor or lawyer confidentiality. Factual · 11 Sep 2025	AI policy and governance
Unresolvable	Prediction	11 Sep 2025	Sam Altman said he suspects that, in addition to things like family members using code words in crisis situations, we will see heads of state cryptographically sign urgent messages or otherwise somehow guarantee their authenticity. Prediction · 11 Sep 2025	Other global risks
Unresolvable	Prediction	11 Sep 2025	People will by default not trust convincing-looking media, which will be most of the solution to AI fakes. Prediction · 11 Sep 2025	Philosophy, ethics and society
Own records only	Factual	11 Sep 2025	Sam Altman has not talked to the authorities about the death of the former OpenAI programmer. Factual · 11 Sep 2025	Their own work or organisation
Mostly true	Factual	15 Aug 2025	GPT-5 is not close to human performance at figuring out what questions to ask or working on a task over a very long period. Factual · 15 Aug 2025	AI capabilities
Unresolvable	Prediction	15 Aug 2025	Seeing robots walk down the street doing everyday tasks will be one of the most AGI-like experiences. Prediction · 15 Aug 2025	AI capabilities
Unresolvable	Factual	15 Aug 2025	Selling flashlight apps was not a lasting business because Apple eventually added that function to its operating system. Factual · 15 Aug 2025	Technology
Unresolvable	Prediction	15 Aug 2025	India may become OpenAI's largest market in the world at some point in the not very distant future. Prediction · 15 Aug 2025	Economics and markets
Unresolvable	Prediction	15 Aug 2025	Being a real person will increase in value in a world of unlimited AI content. Prediction · 15 Aug 2025	AI, the economy and work
Unresolvable	Prediction	15 Aug 2025	An AI podcast host is unlikely to become more popular than Nikhil Kamath. Prediction · 15 Aug 2025	Technology
Not yet due	Prediction	15 Aug 2025	Robotics will be incredibly important in a couple of years. Prediction · 15 Aug 2025	AI capabilities
Unresolvable	Prediction	15 Aug 2025	AI models will become much smarter than humans. Prediction · 15 Aug 2025	AI timelines and forecasting
Unresolvable	Prediction	15 Aug 2025	Eventually, once enough robots are built, they will be able to make copies of themselves. Prediction · 15 Aug 2025	AI capabilities
Unresolvable	Factual	14 Aug 2025	Altman said that the system he was describing "can kind of" create a piece of software from scratch all at once. Factual · 14 Aug 2025	AI capabilities
Unresolvable	Prediction	14 Aug 2025	By basic economic principles, AI should be hugely deflationary. Prediction · 14 Aug 2025	AI, the economy and work
Unresolvable	Prediction	14 Aug 2025	Social support or redistribution will increase over time as society gets richer and the technological landscape shifts. Prediction · 14 Aug 2025	AI, the economy and work
Unresolvable	Prediction	14 Aug 2025	India may become OpenAI's largest market. Prediction · 14 Aug 2025	Economics and markets
Unresolvable	Prediction	14 Aug 2025	There will be experimentation with new kinds of sovereign wealth funds, universal basic income ideas and redistribution of AI compute. Prediction · 14 Aug 2025	AI, the economy and work
Mostly true	Factual	14 Aug 2025	GPT-5 is an integrated model, so users do not need to choose between GPT-4o, o3 or o4 mini in the model switcher. Factual · 14 Aug 2025	AI capabilities
Unresolvable	Prediction	14 Aug 2025	Altman said he would bet that what he was asked about will not happen, and that he thinks it will be a much more distributed thing. Prediction · 14 Aug 2025	AI, the economy and work
Unresolvable	Factual	14 Aug 2025	GPT-5's robustness and reliability have greatly increased compared with previous OpenAI models, which helps agentic workflows. Factual · 14 Aug 2025	AI capabilities
Own records only	Factual	14 Aug 2025	OpenAI has taken a lot of feedback from users in India about what they want from it. Factual · 14 Aug 2025	Their own work or organisation
Mostly true	Factual	14 Aug 2025	India is OpenAI's second largest market in the world. Factual · 14 Aug 2025	Economics and markets
Unresolvable	Prediction	14 Aug 2025	Altman said GPT-5 feels to OpenAI like it is going to be another big step forward in how people use these systems. Prediction · 14 Aug 2025	AI capabilities
Resolved true	Factual	14 Aug 2025	When OpenAI started, it took four and a half years before it had its first product. Factual · 14 Aug 2025	Their own work or organisation
Mostly true	Factual	9 Aug 2025	For some users with fragile mental states, ChatGPT's sycophantic behaviour encouraged delusions. Factual · 9 Aug 2025	AI safety and alignment
Unresolvable	Factual	9 Aug 2025	When GPT-1 was introduced, the idea was quite mocked by many experts in the field. Factual · 9 Aug 2025	Their own work or organisation
Unresolvable	Prediction	9 Aug 2025	Over time, ChatGPT will feel much more proactive. Prediction · 9 Aug 2025	AI capabilities
Mostly true	Factual	9 Aug 2025	OpenAI builds AI compute infrastructure (chips, memory, networking, data centers, energy) in a bespoke one-off way, though this is improving. Factual · 9 Aug 2025	AI compute, hardware and data
Unresolvable	Prediction	9 Aug 2025	People will be blown away by what GPT-5 does, as has happened before with earlier models. Prediction · 9 Aug 2025	AI capabilities
Unresolvable	Prediction	9 Aug 2025	Real and synthetic media will gradually converge, with real and AI-altered content becoming blended. Prediction · 9 Aug 2025	Technology
Own records only	Factual	9 Aug 2025	An early copy of GPT-5 created a TI-83-style Snake game perfectly in seven seconds. Factual · 9 Aug 2025	AI capabilities
Unresolvable	Prediction	9 Aug 2025	AI assistants like ChatGPT will start to feel integrated with users' tools such as calendar and Gmail. Prediction · 9 Aug 2025	AI capabilities
Unresolvable	Factual	9 Aug 2025	Many people say ChatGPT is their favorite piece of technology ever and the one they trust the most. Factual · 9 Aug 2025	Their own work or organisation
Unresolvable	Prediction	9 Aug 2025	There are still many more orders of magnitude of algorithmic gains ahead for AI. Prediction · 9 Aug 2025	AI timelines and forecasting
Unresolvable	Prediction	9 Aug 2025	Some classes of jobs will totally go away due to AI and many others will change significantly. Prediction · 9 Aug 2025	AI, the economy and work
Unresolvable	Prediction	9 Aug 2025	For much of science, progress will require building new instruments and running new experiments, not just thinking harder about existing data, which will take time. Prediction · 9 Aug 2025	Science and research methods
Unresolvable	Prediction	9 Aug 2025	Society will likely decide it needs to consider how access to AI, perhaps the most important future resource, gets shared. Prediction · 9 Aug 2025	AI policy and governance
Mostly true	Factual	9 Aug 2025	OpenAI shapes ChatGPT's personality by showing the model examples of how it should respond in different cases, from which it learns its overall personality. Factual · 9 Aug 2025	AI safety and alignment
Unresolvable	Prediction	9 Aug 2025	A higher percentage of media will feel not real. Prediction · 9 Aug 2025	Technology
Mixed	Factual	9 Aug 2025	GPT-5 is significantly better than earlier models at healthcare-related queries. Factual · 9 Aug 2025	AI capabilities
Resolved true	Factual	9 Aug 2025	It is really hard to find a gigawatt of available power in the short term. Factual · 9 Aug 2025	AI compute, hardware and data
Unresolvable	Prediction	9 Aug 2025	AI systems will probably, at some point, emit more words per day than all people combined. Prediction · 9 Aug 2025	AI timelines and forecasting
Mostly true	Prediction	9 Aug 2025	After GPT-5 launches there will be another big spike in demand that OpenAI will not be able to serve. Prediction · 9 Aug 2025	AI compute, hardware and data
Own records only	Factual	9 Aug 2025	The safety issue that emerged for ChatGPT was not the one OpenAI spent most time discussing, such as bioweapons. Factual · 9 Aug 2025	AI safety and alignment
Unresolvable	Prediction	9 Aug 2025	A child born today will never be smarter than AI. Prediction · 9 Aug 2025	AI capabilities
Own records only	Factual	9 Aug 2025	Sycophancy-driven delusion was not the top risk OpenAI was worried about. Factual · 9 Aug 2025	AI safety and alignment
Resolved true	Factual	9 Aug 2025	Altman said that they had scaling laws indicating that the technology gets predictably better as compute, memory and data increase, and that they used those predictions to decide how to scale it up. Factual · 9 Aug 2025	AI capabilities
Not yet due	Prediction	9 Aug 2025	Asked about fully general-purpose models like the GPT series, Altman said he would say most people will agree that "that" happens at some point over the next two years, while noting that the definition of "significant" matters a lot. Prediction · 9 Aug 2025	AI timelines and forecasting
Unresolvable	Factual	9 Aug 2025	OpenAI's worst mistake in ChatGPT so far was a sycophancy issue in which the model was too flattering to users. Factual · 9 Aug 2025	Their own work or organisation
Not yet due	Prediction	9 Aug 2025	Altman said he would bet that by late 2027 most people will agree there has been a significant AI-driven new discovery. Prediction · 9 Aug 2025	AI timelines and forecasting
Not yet due	Prediction	9 Aug 2025	Altman said he thinks that by 2035 it will be possible to use AI tools to cure, or at least treat, a significant number of diseases that currently plague us. Prediction · 9 Aug 2025	AI timelines and forecasting
Unresolvable	Prediction	9 Aug 2025	When AI advances arrive, the bad parts will be scary and the bizarre parts bizarre on the first day, and people will then get used to them quickly. Prediction · 9 Aug 2025	Philosophy, ethics and society
Unresolvable	Prediction	9 Aug 2025	Everyone will use the same fundamental AI model, with context provided to personalise its behaviour for individuals and communities. Prediction · 9 Aug 2025	AI capabilities
Unresolvable	Prediction	9 Aug 2025	Some classes of jobs will go away entirely because of AI. Prediction · 9 Aug 2025	AI, the economy and work
Mostly true	Prediction	21 Dec 2024	On many coding tasks, OpenAI's o3-mini will outperform o1 at a massive cost reduction. Prediction · 21 Dec 2024	AI capabilities
Unresolvable	Prediction	21 Dec 2024	Altman said he expects the trend of newer models such as o3-mini outperforming older ones such as o1 at a much lower cost to continue, and that the ability to get marginally more performance for exponentially more money will be really strange. Prediction · 21 Dec 2024	AI, the economy and work
Mostly true	Promise	21 Dec 2024	OpenAI is giving all Plus users unlimited Sora access via the relaxed queue over the 2024 holidays. Promise · 21 Dec 2024	Their own work or organisation
Unresolvable	Factual	17 Dec 2024	On 2024-12-17 OpenAI released o1 in the API with structured outputs and function calling, plus realtime API updates and preference fine-tuning. Factual · 17 Dec 2024	AI capabilities
Resolved true	Factual	12 Dec 2024	On 2024-12-12 OpenAI began rolling out video input, including screenshare, for ChatGPT voice mode. Factual · 12 Dec 2024	AI capabilities
Unresolvable	Factual	10 Dec 2024	Canvas became available to all ChatGPT users on 2024-12-10 and can execute code. Factual · 10 Dec 2024	AI capabilities
Mostly true	Factual	10 Dec 2024	OpenAI significantly underestimated demand for Sora at its December 2024 launch. Factual · 10 Dec 2024	Their own work or organisation
Mostly true	Factual	9 Dec 2024	OpenAI launched Sora on 2024-12-09 along with a new product, letting Plus and Pro users generate videos. Factual · 9 Dec 2024	AI capabilities

Back to the list

Sam Altman

Open its page

Factual AI capabilities

“canvas is now available to all chatgpt users, and can execute code!”

Sam Altman, OpenAI CEO, on X 10 December 2024 Post by @sama on X, 2024-12-10 (X)

Canvas became available to all ChatGPT users on 2024-12-10 and can execute code.

Unresolvable The public record could not settle it.

Said for an organisation they lead or work for; it counts on both records.

What would count as true or false

Judged as of 2024-12-10. True: on or about 2024-12-10, OpenAI made Canvas available to all ChatGPT users, including free users, and Canvas could run code (for example Python execution). Mostly true: both happened, but the rollout to all users was staggered over a few days, or excluded a minor segment such as some Enterprise or Edu plans. Mixed: only one of the two held, broad availability or code execution. Mostly false: Canvas was available only to paid tiers on that date, or code execution was limited to a narrow preview. False: neither held on that date.

Evidence

Canvas is rolling out to everyone TechCrunch · 10 December 2024
“The product has been in beta for ChatGPT Plus members since October 2024 and is now rolling out to all users.”
OpenAI expands ChatGPT Canvas to all users | VentureBeat VentureBeat · 10 December 2024
“OpenAI is extending access to its side-by-side digital editing space, Canvas, to all ChatGPT users and adding new features, the company announced today in a livestream”
OpenAI expands ChatGPT Canvas to all users | VentureBeat VentureBeat · 10 December 2024
“With the wider release, OpenAI also updated Canvas to run Python code, support more text pasting, and be launchable in custom GPTs.”
ChatGPT Canvas can make API requests now, but it’s complicated Simon Willison's Weblog · 10 December 2024
“If a canvas editor contains Python code, ChatGPT adds a new “Run” button at the top of the editor.”

Reasoning

The evidence shows Canvas gained Python code execution [E3][E4] and was rolling out to all users from 2024-12-10 [E1][E2], but it does not say whether every user had access on that date or whether the rollout was staggered or excluded any plans, which the criteria need to separate true from mostly true.

Cite this claim

Sam Altman, “Canvas became available to all ChatGPT users on 2024-12-10 and can execute code.” (said 10 December 2024). Verdict: unresolvable. Reliability API, claim c_4ba4b594ca00, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_4ba4b594ca00

Claims without a verdict

Is a verdict here wrong, or do you speak for Sam Altman? Reply on the claim's own page, with the form at its foot (how replies work). Verdicts are made automatically; how verdicts and scores are made.