Smart contract is law is possible only if you have perfect smart contracts. It's the immutable idea which is the problem because in order to have law in the sense that humans think of it you need the ability to reinterpret the law.
This is why I never liked to call it "law" and why I called it "rules". Rules are much more clear because rules are based on formal logic. The problem with the imperative Turing complete smart contract language is that you don't have consistent logic. This might not be the only problem but myself and others have identified that it is the primary problem.
To illustrate you can take propositional logic and when it's consistent you will always know that in the end there is either a TRUE or a FALSE. There is no TRUE and FALSE because that would be a contradiction and logically impossible. This shows that at the core of security of any language is the logic itself because only through the logic can you control for unexpected behaviors and even predict in advance all the possible behaviors without having to run through the smart contract to find out.
The idea is to minimize the risks of financial loss by being extremely careful and formal about writing the code for the smart contracts which manage potentially millions or billions of dollars.
References
https://en.wikipedia.org/wiki/Propositional_calculus