Nope, it just has to be shown to be relevant to the requirements of the job.
So you call it an "employment skills assessment," make 80% of it indistinguishable from an IQ test, and add some domain-specific questions related to the industry or the position.
P&G has done this for a long time, among other companies.
As best I can tell 'Judge Wood ruled that an older state-certification test, which was intended to measure teachers’ knowledge of the liberal arts and science, was racially discriminatory.'
Elsewhere I'm seeing "The court found that Black and Latino teachers clearly passed these tests at lower rates than white teachers. In order to prove that this wasn't illegal, the defendants had to show that the test actually demonstrated what it promised: that teachers who did well on the test would do better in their jobs."
It seems to me that the error was using the wrong test and it was the wrong test because it had disparate impact. What it does not appear to be is a case that is not about disparate impact because it instead had errors.
If men get into traffic accidents more often than women, then clearly there is something wrong with the cars and car manufacturers should be sued for their discrimination?
In the Eastern bloc, agriculture was immensely held back because everybody had to follow the ideology and adjust their science and methods to "I god damned said so!" of the rulers. Seems like America is curious about following the same path.
(I'm writing as if I think real IQ tests are a good idea, and they are not --- in fact, that's my whole point: there's a mythology that IQ tests aren't used because they're illegal, but they are not that; what they are is ineffective.)