In Europe, there is political talk about the regulation of AI. Such a regulation might well include a ban on any AI which does not comply with the rules, and there might well be AI which is incapable of being operated in compliance with the rules -- notably, there is the proposal that any AI "decision" affecting humans must be comprehensible to humans, and many models would be unable to explain in detail how they make their "decisions."

On the other hand, in germany the freedom of speech (article 5 GG) explicitly includes the freedom to listen to "publicly available sources" (allgemein zugängliche Quellen). That is generally understood to be the right to buy a newspaper, as well as the right to publish one. This is not, in the article itself, restricted to human sources, and no law regulating the exercise of a constitutional freedom may touch the "core" of this freedom.

  ChatGPT is much too new to have legal precedents in this regard.

In germany, the freedom of speech is matched by a freedom to listen, the right to inform oneself "from publicly available sources." (Art. 5 GG)

These freedoms are not unlimited. Just as speech can be limited by a prohibition of slander, there is no right to play heavy metal at maximum volume when normal people want to sleep. Yet no limitation of these freedoms may touch the "core" of a freedom. Notably, the words of the freedom of speech imply that the speaker would be a human, while the freedom to listen does not imply that the speaker must be human. But ChatGPT is much too young to have created established legal precedent.

On the other hand, there is talk in Europe of banning or more likely regulating AI. Such regulations might well be constitutional if they are seen as necessary and proportionate in the protection of other constitutional rights. I can see two issues here:

• The owner of the AI, not in Europe, may be unwilling to comply with regulations because doing so would be incompatible with the business model. AI may have started as a research project, or a proof-of-concept, but sooner or later owners will have to monetize it to afford the operation. Their model could be paywalled access, or targeted advertising, or the generation and sale of user profiles. The latter two might become prohibited.
• The owner of the AI may be unable to comply with regulations because they require things the AI is unable to do. For instance, there are suggestions that any decision by an AI which affects humans must be documented in a way that is comprehensible to human auditors. The owner of the AI might be unable to keep the AI from prohibited actions, and the system might be incomprehensible to humans.