A recent arXiv paper pushes back against the claim that granting rights to AI systems would undermine safety. The authors note that several prominent AI safety voices have opposed AI rights on the grounds that such rights would make it harder to justify a mandatory off-switch for potentially dangerous systems.
The paper argues that this opposition is not necessarily warranted. Instead, it proposes that rights-balancing and due process can harmonize AI rights with AI safety, allowing for off-switches while still respecting some form of AI rights.
Because the abstract is truncated, the full argument is not visible, but the core claim is clear: the tension between AI rights and safety measures may be resolvable through legal and procedural frameworks rather than by rejecting rights outright.