There is a strategy in AI defense mechanisms. It's a crawling denial of service attack. When the AI is, let's say, mining data, it doesn't know that the content that was captured or is currently being captured is a never ending story (because it's source is AI on the other end). At a certain point the crawling AI is lacking entropy (as a mitigation) in the data source, but the delay objective (denial of service) attack was a success. In machine learning terms, it finds a local minimum in gradient decent. If the warzone/environment is littered with these strategicly placed dead ends, then you loose the race over AI crawlers that placed the landmines.
Remember that we live in a world with generative AI. We NEED a pinch test, in watts, which AI can't do yet. What kind of strategy have you developed that 'touched ground'?
