"AI models are created in the West, therefore they may reproduce Western assumptions."

There is a reasonable concern behind this.

AI systems are trained primarily on material that exists in the published record.

If a view is underrepresented in published scholarship, AI systems may also underrepresent it.

However, that does not automatically mean the neglected view is correct.

The question remains:

What evidence supports the view?

rather than

How popular is the view?

Popularity and truth are not identical.