Perhaps this is one of the understanding gaps that crop up around AI development? At my current company and most others I've worked at, testing capability is part of the same bucket because engineers do their own QA.