Ask for a poem, a workout plan, or an explanation of a difficult passage, and ChatGPT can reply in conversation. Released by OpenAI on November 30, 2022, it is a chatbot accessed through a web interface. Its public launch brought large language models—AI systems that generate and work with language—to a much wider audience. It reached 1 million users within days and 100 million within two months. section 1.0 section 1.16
ChatGPT is useful for drafting, summarizing, answering questions, and revising text. These abilities can make it feel knowledgeable, but a fluent answer is not necessarily a correct one. Large language models can invent information, reproduce bias, and impose environmental costs. Treat consequential responses as drafts or leads to check, not as verified facts. section 1.0 section 1.22 section 5.8
A convincing but false legal brief
In Mata v. Avianca, an attorney used ChatGPT to prepare a court brief that cited cases that had never occurred. He later said he had not realized its output could be false. The episode shows why plausible writing needs checking against reliable sources, particularly in legal, medical, and financial work. section 5.8
A helpful draft with a privacy cost
An attorney who pastes a client’s contract into ChatGPT for suggested revisions may save time while exposing confidential information. Prompts can contain sensitive details, and their handling can create privacy or professional obligations. A useful answer does not remove the need to decide what is safe to share. section 3.10
The trade-off also appears in healthcare. ChatGPT produced a patient-facing script about alcohol use disorder and rewrote it in more accessible language. Such help with wording may be valuable, but an empathetic tone should not be mistaken for sound medical advice or a replacement for human care. ChatGPT’s success helped accelerate investment and competition in generative AI; how to use these systems responsibly remains a practical question. section 6.2 section 1.22
