| # FAQ – OpenAI Retired Model Proposal | |
| --- | |
| ### ❓ Why should OpenAI release GPT‑3 or GPT‑4? | |
| These models are **no longer used in OpenAI’s primary commercial systems**. Open-sourcing them would: | |
| - Support global education and research | |
| - Improve transparency and reproducibility | |
| - Enable low-resource developers to build offline and secure AI applications | |
| --- | |
| ### ❓ Isn’t this dangerous? What about disinformation? | |
| Other major models (LLaMA, Mistral, GPT-J) are already open. Safety is a matter of **context and governance**, not secrecy. Openness allows third-party alignment work to flourish. | |
| --- | |
| ### ❓ How would OpenAI protect itself legally? | |
| The **RMLA (Retired Model License Agreement)**: | |
| - Disclaims liability | |
| - Disallows misuse of OpenAI branding | |
| - Makes usage and risk solely the responsibility of the end user | |
| --- | |
| ### ❓ Would this hurt OpenAI’s business? | |
| No. GPT‑4 is already being replaced by GPT‑5. These legacy models are outperformed and no longer central to OpenAI’s revenue stream. | |
| --- | |
| ### ❓ What about hardware damage or hallucinations? | |
| The RMLA waives OpenAI’s responsibility for: | |
| - Hardware wear due to intense workloads | |
| - Misinformation or hallucinated content | |
| - Any real-world harm caused by derivative uses | |