"Read contract language for what it permits, not what it promises When Anthropic's CEO Dario Amodei rejected the Pentagon's revised contract terms, he said the proposed concessions came with 'legal ease that would have made them ineffective' — language that sounded protective but would have allowed the very uses Anthropic was trying to prohibit. Kevin Roose noted this is distinct from a political vendetta: 'OpenAI is sort of using this sort of legal ease to frame this as a victory when really they have conceded to the thing that Anthropic objected to.' Before signing any agreement with meaningful ethical stakes, scrutinize not just what the contract prohibits but what it still permits through omissions and definitional loopholes. Source: Kevin Roose and Casey Newton, Hard Fork, 'Anthropic vs. the Pentagon'"