`String--derived models and light U(1) embeddings' Extra U(1) extensions of the Standard Model arising from string theory have been considered in the literature since the mid-80s. Such a U(1) can play an important phenomenological role in solving the mu problem and suppressing other unwanted operators, such as dimension four and five proton decay mediating operators. Yet keeping the U(1) unbroken down to low scales in quasi-realistic string models turns out to be highly non-trivial. In this talk I will discuss the constraints and some scenarios in which they may be alleviated.