AI threatens democracy so govt should regulate it

As the AI revolution gathers pace, a growing number of voices are warning of the threats it poses to democracy. There is one obvious way governments can limit the damage. If an AI has only a single purpose such as winning a debate, writing a threatening or disinformation letter or tracing e-mail addresses, then limiting its generality to a specific task or domain can limit its ability to subvert democracy. For example, there is no evidence that AI article writers are capable of writing anything original - they are only capable of re-expressing what is already in the training data. So an AI can be limited to this task by allowing it to only express things it already knows (i.e., the training data). An AI for political debate can be limited to debating a specific topic like say gun regulation in the US or immigration in Europe. If the AI is a chatbot having a conversation with a user, then the chatbot should be unable to change the topic of conversation. A Chinese AI tracing the email addresses of candidates for the CCP can only trace the email addresses of known candidates. A Chinese AI writing disinformation letters to Chinese citizens can only express things it already knows.