OpenAI releases ChatGPT as a public research preview

OpenAI released ChatGPT as a free public research preview on 30 November 2022. The web interface allowed users to converse with a model trained to follow instructions, answer follow-up questions, acknowledge some mistakes and reject certain unsafe requests.

The initial system was based on models in the GPT-3.5 series and further trained using human demonstrations and preference feedback. It generated text from statistical patterns rather than consulting a guaranteed database of truth. OpenAI warned that outputs could be plausible but incorrect, biased or sensitive to prompt wording.

Public access rapidly expanded awareness of generative artificial intelligence beyond specialist and developer communities. People tested it for writing, programming, education and search-like questions. Those uses demonstrated versatility but did not establish reliability or permission to replace professional judgment in high-stakes settings.

ChatGPT was a product interface and continuing service, not the same entity as the underlying model family. Later subscriptions, GPT-4 integration, multimodal features and policy changes did not exist in the release in their eventual form and should not be read backward into the 2022 preview.

OpenAI's launch post establishes the date, purpose and stated limitations. The release-day account captures the opening of broad public access, not the invention of chatbots or large language models. Its historical significance comes from adoption and visibility, while claims about productivity, labour or social impact require subsequent empirical evidence.