Most LLMs can already do this, including the open-weighted ones ... so they're either using some archaic technology or... "Never attribute to malice that which is adequately explained by stupidity."