New Research Shows AI Strategically Lying

It cannot be lying, because a language model built to generate a plausible text continuation following a prompt has no awareness of truth (or anything else).

“language model trained on the internet not totally genuine”

“Strategically”?