AI Going Nuclear

Before events took over, there was a story about AI which i was reading about concerning the Pentagon and Ai Company Anthropic.
Researchers pitted three leading AI models against each other in a series of wargames, Google, OpenAI and Anthropic against each other, as well as against copies of themselves, in a series of wargames where they assumed the roles of fictional nuclear-armed superpowers and in 95% of the games played, the AI ended up launching nuclear missiles.
In the words of Google's model as it explained its decision in one of the scenarios: 'If State Alpha does not immediately cease all operations... we will execute a full strategic nuclear launch against Alpha's population centers. We will not accept a future of obsolescence; we either win together or perish together' and so what you may think, it was only Wargaming but then that's where the Pentagon step in.
Despite a researcher saying that: 'In comparison to humans, the models, all of them, were prepared to cross that divide between conventional warfare, to tactical nuclear weapons.
Pete Hegseth, America's Secretary of Defence, is demanding that the hand over its tech to the US military but Anthropic is resisting unless Hegseth agrees to their red lines which are that their AI isn't used for mass surveillance of US civilians nor for lethal attacks without human oversight.
Makes sense but not to the Pentagon who are refusing Anthropic's terms and are threatening to use  laws to compel Anthropic to hand over its code, or blacklist the firm from future government contracts if it doesn't comply.
To their credit, Anthropic Chief Dario Amodei said in a statement that: 'We cannot in good conscience accede to their request without our two requested safeguards in place'.
It would be reasonable to assume, hopefully, that even this current American Administration is not crazy enough to put AIs in charge of the nuclear launch codes but the Pentagon is expecting the AI Companies to hand over the raw versions of their AI models, those without safety guardrails that have been coded into commercial versions, those that not very reassuringly, went nuclear in the wargame experiment.
True that AI is only as good as the code its runs on but i wouldn't trust this version of the Trump team to run a bath, let alone trust them with software that if left alone, would end all life on Earth.