They can have cultural or social biases unless very carefully calibrated. Ideally you should be able to take anyone who speaks a reasonable level of English, give them test and have them get a score that matches their intelligence percentile.

In practice, the tests are built for a particular socioeconomic class and type of person. As a trivial example, imagine the test uses the word "yacht" and you give the test to someone who has never culturally been exposed to that word - that person will, on average, score a bit lower than someone who recognizes the word immediately.

Or imagine you give a test using Imperial units to a European who exclusively grew up using Metric (or vice versa). Knowing that three feet are in a yard isn't an intelligence test, it's a test of cultural familiarity.