Data hazards occur in pipelined processor architectures when instructions that exhibit data dependencies are executed out of order, leading to incorrect results. They can be mitigated through techniques such as forwarding, stalling, and reordering of instructions to ensure data integrity and correct program execution.