From securityboulevard.com
A safe language is a language which protects its abstractions. For instance, that the run-time cannot be corrupted by writing past the end of some memory region, that uninitialized memory is not read, that lexically scoped variables cannot be changed outside their scope, and so on. A statically checked language is a language which ensures the safety of its abstractions at compile-time. For instance, by checking that all variables are initialized before they are used. A safe language, which is dynamically checked, must enforce these safety properties by performing checks at run-time.