After all, we're still not 100% sure how LLMs make their decisions in what they string together as output, so the company's not _technically_ lying.