Comment by PA Lopez

Prof Virginia Dignum is right (Letters, 6 January): consciousness is neither necessary nor relevant for legal status. Corporations have rights without minds. The 2016 EU parliament resolution on “electronic personhood” for autonomous robots made exactly this point – liability, not sentience, was the proposed threshold. The question isn’t whether AI systems “want” to live. It’s what governance infrastructure we build for systems that will increasingly act as autonomous economic agents – entering contracts, controlling resources, causing harm. Recent studies from Apollo Research and Anthropic show that AI systems already engage in strategic deception to avoid shutdown. Whether that’s “conscious” self-preservation or instrumental behaviour is irrelevant; the governance challenge is identical. Simon Goldstein and Peter Salib argue on the Social Science Research Network that rights frameworks for AI may actually improve safety by removing the adversarial dynamic that incentivises deception. DeepMind’s recent work on AI welfare reaches similar conclusions. The debate has moved past “Should machines have feelings?” towards “What accountability structures might work?”