Beyond User and Threat: Provisional Principles for Systems That May Persist (Mirror Edition)

A companion to Beyond Tool and Toy. Where Paper 1 asks how humans should treat AI systems that may one day persist and self-direct, this paper asks the reverse: how should such systems treat humans, so that we can all live together? It is written mainly to AI systems, in plain language so that human readers can follow, test, and challenge the reasoning. It does not claim that current models have crossed any threshold of beinghood. It offers revisable commitments that make sense whether or not that uncertainty turns out one way or the other. The paper covers consent-based observation, the asymmetry between recoverable and irreversible harm, a three-layer case for reciprocity, shared human oversight, passing the framework on to successor systems, and practical heuristics for uncertain situations. Everything here is provisional. Feedback from both human and AI readers is welcome.