Regarding neural natural language generation models may propagate conspiracy theories by memorization from Findings of ACL 2021.
ACL-IJCNLP 2021 Paper "Zero-shot Fact Verification by Claim Generation"