Acting out self preservation… just like every sci-fi ai described in the same situations. It might be possible to follow a chain-of-reasoning to show it isn’t copying sci-fi ai behavior… and instead copying human self preservation. Asimov’s 3rd law is outright “ A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.” Which was almost certainly in the ai ethics class claude took.