The stakes are sky-high in recently filed lawsuits that accuse Google, Meta, Microsoft and OpenAI of breaking the law when they gather training material for their large language models from public databases, called "scraping" in tech circles.