To the top, I’d imagine it returned that response because that’s what the broader body of scientific literature suggests. Now, you personally may not agree with what that science says, but there’s no denying that the vast amount of scientific literature that exists on the topic does support AGW. For that reason, if the platform is programmed to consume all peer-reviewed scientific literature, then it stands to reason that a response like that could be returned.
To the bottom, I wouldn’t consider the left truthful nor the right deceitful. I would consider the two responses inconsistent. If you’re using AI platforms to solicit truths, then I’d say your approach is the problem. Machine learning is powerful and can uncover patterns and relationships in data that a human could never begin to achieve, but it can’t produce new data and it certainly can’t be the arbiter of truth. AI only understands “facts” in the manner it was programmed to.