Update locks in SQL Server are used to prevent deadlocks that could occur when multiple transactions try to concurrently update the same resource. Without update locks, transactions would first acquire shared locks before updating data. However, this could lead to deadlocks if two transactions simultaneously tried to convert their shared locks to exclusive locks in order to perform an update. To avoid this, SQL Server uses update locks, which are incompatible with other locks and ensure that only one transaction can update a resource at a time by first acquiring an update lock before converting it to an exclusive lock.