How should AI be regulated, if at all?

While the advantages that AI can bring are numerous and obvious, there are also many ways in which this powerful technology can have negative impacts: e.g. by violating our privacy, making us addicted, stealing our jobs, misleading us with fake news, discriminating against disadvantaged demographics, increasing inequality, and perhaps even exterminating us all. Based on fears like these, some people have called for regulations on the use, development and sometimes even the research of artificial intelligence. Of course, some regulations already exist in the form of laws that were on the books, sometimes long before computers even existed, which assert rules and rights for people that others must respect (with or without AI).

Of course, the meaning of these laws and how to apply them can be difficult to interpret, and many feel they should be updated to more clearly reflect the current and future technological landscape. But should the new laws be more restrictive in order to combat misuse and prevent the invention of apocalyptic technology, to avoid the negative impacts listed above? Or should they be more permissive, so as to avoid stifling innovation? Can we trust companies and individuals to behave ethically without explicit regulations? Should regulations be implemented by individual national governments? A group of governments (e.g. the UN)? The academic community? Industry coalitions? Is it enough to regulate (mis)use of this technology, or will it also be necessary to regulate the research itself? What should the regulations look like, and who should decide?