assuming it's "aligned" with a foundational model lab