Coming soon: No personhood for AI in Missouri

Missouri's AI Non Sentience and Responsibility Act gets the present right, and that matters because clarity today prevents confusion tomorrow. AI is not a person, AI is not conscious, and AI cannot marry, furthermore, AI can't sit on a board, and AI can't own property. Responsibility is on humans, where it belongs, because we treat present day systems as tools and we prepare frameworks for the future only when verifiable evidence arrives. No current AI meets scientifically defined thresholds of autonomy or awareness, therefore personhood is human only, and liability is human only.

Getting the present right is only half the task, however, the bill's core strength is that it blocks liability shell games and stresses human oversight. Owners are liable for how they steer and deploy AI, and developers and manufacturers are liable for defects and insufficient instructions, moreover, courts are free to pierce the corporate veil when companies intentionally undercapitalize AI subsidiaries, or hide control behind layers of shell companies. That is healthy governance, because it rewards real risk management and punishes theater, and it is entirely consistent with our demand that alignment cannot be a marketing label, and cannot be claimed without measurement and continuous verification.

We must also be concrete about what responsible deployment looks like in practice, nevertheless, for any system with impact on health, safety, property, or public welfare, we demand audited logs, escalation paths, and regular risk assessments, and we demand that those assessments are tied to clear thresholds that trigger human intervention. We need tiered trust, and we need the most impactful decisions to require more evidence, more cross-checks, and explicit human signoff, furthermore, we need independent assessments to verify whether oversight is for real or for show. These are not burdens on innovation, however, these are the rails that keep innovation honest.

We need cultural safeguards to prevent people from outsourcing moral agency to software, because the desire to say the AI made me do it is a potent one, and it grows as the systems get more conversational and more authoritative, therefore, we need to counteract that desire with governance and practice. The JULIA Test is one simple way to measure drift into unhealthy patterns, and it asks if a person is attributing feelings or emotions to the model, seeking approval from it, or concealing dependence, or if a person is becoming uncomfortable making decisions without it. The purpose is not to stigmatize usage, however, the purpose is to maintain human autonomy and accountability.

Others will argue that a strong non sentience declaration will foreclose future debate, although it will not, because it keeps us grounded in reality now while leaving the door open to evidence later. If one day a system meets robust, peer reviewed criteria of consciousness, then we reassess, and we do so with cool criteria, not hype, because rights do not precede capacity, they follow from it, consequently, until then the appropriate stance is partnership with tools, not parity with people.

In business and government, the way forward is pragmatic, therefore, we consider AI as a controlled instrument and demand that owners publish risk inventories for high stakes use, define failure modes and monitoring thresholds, and perform red team exercises and cross validation between diverse models for critical decisions. We teach teams to interrogate recommendations, not rubber stamp them, and we log interventions and near misses, so that learning compounds, moreover, these habits not only reduce harm, but also improve performance, since they bring blind spots and brittle reasoning to the surface before they become incidents.

In summary, Missouri's bill is a bright line that works for everyone, because humans have personhood and responsibility, and AI has neither, furthermore, marketing cannot launder liability, and this is exactly the foundation we need to build safe and effective systems in the present. To that we add verifiable safeguards, tiered reliance, and culture that refuses to abdicate its judgment, and if the science ever changes we will be prepared to review that with rigor and humility, consequently, in the meantime we govern resolutely, we measure truthfully, and we prefer coexistence through accountability over control theater and speculative hyperbole.