Global publishers are increasingly challenging artificial intelligence developers using copyrighted material without permission for large language model training under fair-use arguments