Bob is selling a vehicle to Alice. He asks ChatGPT to write a contract, and it does, for some reason, write it. (It has filters in place to prevent users from writing legal documents, but these filters might be possible to bypass.) They both skim over the contract and sign it; the document they sign has the exact text from GPT with no changes or additions.
Unsurprisingly, given the nature of current generative AI technology, there is an ambiguity in the text. This materially affects the transaction, and they cannot settle the dispute themselves, so they to to court. If Bob had written it manually, the court would likely resolve the ambiguity in Alice's favor. Is that still the case when he had an AI write it? If not, would it be the case if he had slightly edited the AI output, but the ambiguity were in a part that the AI wrote?