Not a trust score: it covers only the claims we've checked, and small samples move a lot. Last PixProof Sep 22, 2026.
Quick facts
- What is OpenAI?
- OpenAI is an American artificial intelligence company that develops large language models, including the GPT series and ChatGPT.
- Where is OpenAI from?
- United States. Headquartered in San Francisco, California, U.S.
- What is OpenAI's full company name?
- OpenAI Inc.
- When was OpenAI founded?
- In 2015, by Sam Altman, Elon Musk, Ilya Sutskever and Greg Brockman.
- How many people work at OpenAI?
- About 4,500 (as of 2026).
- Where does the name "OpenAI" come from?
- It's named after openness and artificial intelligence.
▸ ABOUT OPENAI
OpenAI is an American artificial intelligence company that develops large language models, including the GPT series and ChatGPT. Founded as a nonprofit AI research lab, it restructured into a hybrid for-profit/nonprofit structure and, as of 2025, a public benefit corporation overseen by the nonprofit OpenAI Foundation.
▸ WHO OWNS WHAT
From Wikidata (see record). Lists can include companies since merged or closed.
▸ TRACK RECORD
OpenAI's history and the 3 marketing claims we've checked, in date order — so you can see what they promised when, and how it held up.
- 2015Founded on December 8 as a nonprofit AI research lab in San Francisco.
- 2019Restructured to add a 'capped-profit' arm, OpenAI LP, to attract investment while keeping the nonprofit mission.
- 2022Launched ChatGPT, which drove mass public adoption of consumer AI chatbots.
- 2025Restructured again into a public benefit corporation overseen by the nonprofit OpenAI Foundation.
- Sep 9, 2026CLAIM · OpenAI's Navier-Stokes "Proof""OpenAI's model autonomously produced a genuine mathematical proof relevant to the Navier-Stokes singularity question."→ 🟡 QUESTION IT Read the PixProof →
- Checked Sep 2026CLAIM · GPT-6 Astra"GPT-6 Astra beats Anthropic's Claude Fable 5.1"→ 🔴 BYE THE HYPE Read the PixProof →
- Checked Sep 2026CLAIM · OpenAI's "Safely Sandboxed" Agent Evaluations"Capability evaluations of frontier AI agents are run in sandboxes that prevent real-world impact even if a model behaves in unexpected or risky ways."→ 🔴 BYE THE HYPE Read the PixProof →
Claim dates come from when the claim's own source was published; "Checked" means we couldn't date the claim, so it sits at the date we verified it.
