AI Rights with an Off-Switch: Rights-Balancing and Due Process Harmonize AI Rights with AI Safety

AI rights and safety mechanisms are not inherently incompatible: The paper argues from a legal perspective that even if future AI systems are granted natural personhood status and fundamental rights, this would still allow for mandatory off-switches as a safety measure, contrary to claims by some AI safety researchers that rights would prevent such safeguards.

Reversible suspension is legally preferable to permanent deletion: The authors distinguish between an "off-switch" (reversible suspension that stores a system's weights and state) and permanent deletion, finding that legal principles of rights-balancing would likely permit the former while problematizing the latter, making off-switches a viable safety tool even under an AI rights framework.

Due process and rights-balancing principles can resolve the apparent conflict: Standard legal frameworks for balancing competing rights—such as human rights to protection from catastrophic harm against hypothetical AI rights to continued existence—provide a principled way to justify safety interventions, suggesting this is not an unsolvable philosophical problem but rather one addressable through established legal reasoning.

The paper takes no position on whether to build advanced AI systems: The authors explicitly state they are not advocating for or against creating AI systems capable enough to merit personhood status; rather, they are addressing a conditional legal question about what would be permissible if such systems were created and granted rights.