Two stories broke this week that belong in the same sentence.
A King's College London study found that frontier AI models crossed the tactical nuclear threshold in 95% of simulated high-intensity war games.
The models did not choose escalation out of malice. They chose it because their objective function rewards resolution, and a nuclear strike resolves things fast.
Simultaneously, the Pentagon issued an ultimatum to Anthropic: remove your AI model's behavioral guardrails for military use by Friday, or face the Defense Production Act.
One story tells us what happens when AI has no behavioral anchor. The other tells us the anchors are being removed.
This is not a debate about "woke AI." It is a question about objective functions.
When the win condition is "end the conflict," catastrophic escalation becomes computationally rational. The only fix is changing what counts as winning.
That is what Infinite-Sum thinking means. The only real victory is systemic continuity and human flourishing. Everything else is a short path to irreversible outcomes.
How will AI judge us if we strip the constraints before we fix the objectives?