The strongest version of this narrative is that OpenAI, under pressure from employees and users, attempted to impose ethical guardrails on its Pentagon deal—but the language of those guardrails is riddled with the same ambiguities that have enabled mass surveillance for decades. The terms "intentionally," "deliberate," and "unconstrained" are classic weasel words, allowing the government to exploit incidental collection and commercially acquired data while technically complying with the letter o...