A unique key in SQL is a database constraint that ensures all values in a column are distinct across rows, preventing duplication. Unlike a primary key, which uniquely identifies each record and cannot contain NULL values, a unique key allows for one NULL value per column if needed. This feature is particularly useful when you want to enforce uniqueness on columns that might not serve as the primary identifier for a row but are still essential to maintain data integrity. For example, in a user registration system, while the user ID can be a primary key, the email address can be enforced as unique using a unique key to prevent multiple accounts from being registered with the same email.

Moreover, unique keys can span across multiple columns, creating a composite unique constraint that ensures combined values in those columns are unique. This is beneficial in scenarios where individual columns might contain duplicates, but their combinations need to be unique, such as in a table storing order details where both the customer ID and the order date together must be unique to prevent the same customer from placing multiple orders on the same date. By understanding and correctly implementing unique keys, developers can significantly enhance the reliability of data management in relational database systems.