I probably don't understand it but, the reports are all that it is Anthropic, not OpenAI, that is the one supplying the AI fuel for the present Iran war. (Thus the delay in the imagined military rejection of Anthropic). Meanwhile OpenAI says it has exactly the same principles and 'red lines' but is somehow implementing them differently. They seem to agree that they do not differ in any of the respects in question. The real story has to be completely different.