Some of it was spin, but some of it (like the clinical trials and automotive safety system design, thinking non-native English speakers are "fake") is definitely caused by a lack of information due to the subset involved.
lol imagine the outrage if a title like this was about any other race or sex. Some poor "journalism" from nostr:npub1va6zy0gmr8wp2328xw4h2kdvmcwjrl0nl6djv5kyv0rl4kcjcppqvzckg7
https://techcrunch.com/2024/05/22/metas-new-ai-council-is-comprised-entirely-of-white-men/
Discussion
That isn't to say that a panel of white, American men wouldn't have done an excellent job, but then you'd have to give preference to the ones with more empathy and diverse life experiences / exposure.
I don't need to be a non-native English speaker or someone with a non-American accent to recognize that such people should be accounted for. Massive oversight in the training data and nobody caught it for years. All levels failed on the QC (and I doubt all levels were only white, American men).
Or they're white, American men, but ignorant or obsessively woke, so you end up with black Vikings and female Aboriginal popes.
This is common in any field involving designing algorithms, tools, or rules/laws.
The group of people effected by or using the end product is often very different from the people building the product, and they don't do effective market research or quality management to make up for that, so the result misses the mark.
It's also not an answer to just "have hiring diversity" because, for example, having a woman on the team doesn't mean they will be any good at empathizing with other women. Maybe they just have a political ax to grind or are socially inept, so adding them to the team makes the result worse, instead of better.