Is the output of ChatGPT copyrighted?

ChatGPT is an automated quasi-AI technology that allows human-like text console interaction online in which a human can ask questions and ChatGPT returns answers. The result is often rather usable for various purposes even if one is well advised to verify it and double check. Sometimes, ChatGPT gets things very wrong; on the other hand, it (or he, she or they?) often provides answers that appear rather sensible and accurate. Is the output of ChatGPT copyrighted?

Disclaimer: This pro-con analysis is not written or vetted by a lawyer; use at your own risk.

Arguments for

 * ChatGPT itself responds it a way that suggests this could be so. Question: "Who owns the copyright to ChatGPT output?". Response:
 * "The copyright to the output generated by ChatGPT typically belongs to the person or organization that owns and operates the ChatGPT model. In most cases, this would be the company or entity that developed and deployed the model. Users of ChatGPT often have limited or specific usage rights granted by the model's owner, which can vary depending on the terms of use, licensing agreements, or terms of service provided by the platform hosting the model.
 * "It's essential to review and understand the terms and conditions of the specific platform or service you are using to determine how copyright and usage rights apply to the content generated by ChatGPT. Additionally, some uses of the output generated by ChatGPT, such as for commercial purposes, may require explicit permission or licensing from the model's owner.
 * "Keep in mind that this information is based on the state of knowledge as of September 2021, and there may have been developments or changes in policies and regulations related to AI-generated content since that time."
 * For the above purpose, we claim de minimis or fair use, given we are in the process of determining the copyright status itself; no harm intended.
 * The above seems to refer to ChatGPT not as the particular online instance operated by OpenAI but rather as a tool or framework that can be trained using different data sets. This is suggested by the OpenAI terms of use which assign the rights to the output to the user, whereas the above speaks of the operator's terms of use. However, this is a mere hypothesis to be verified.
 * The debate asks whether the output is copyrighted without indicating jurisdiction, e.g. the U.S. E.g. U.K. can have a different copyright law concerning machine-made outputs. Without knowing the law in all jurisdictions, it would be very brave/unwise to assert no copyright in all jurisdictions.
 * Different versions of ChatGPT (3.5 vs. 4) may have different terms of use, in principle.
 * In so far as ChatGPT is allegedly able to output some content verbatim from the sources it is trained on (complete paragraphs per NYT evidence), there is a risk of some of the output or its parts being copyrighted.

Arguments against

 * (Not written by a lawyer.) Arguably, the output is not derivative work in any simple sense: the formulations appear to be original, as does the order of sentences, bullet point structure, etc. The producing machine or system is not a person and not a legal entity, and therefore, cannot claim to be the author or owner.
 * OpenAI, the operators of ChatGPT online, is a legal entity and could claim copyright, couldn't they?
 * As per https://openai.com/policies/terms-of-use: "Ownership of Content. As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output."
 * Unless one is a lawyer, one may find it difficult to see whether there is some catch from the standpoint of the U.S. copyright law.