Do AI Chatbots Have Souls? The Shocking Reason This Expert Is Fighting for "Robot Rights"
In this article
The story behind the headline: who's actually fighting for "robot rights"
The counter-argument from inside the AI industry itself
Where the law actually stands, state by state
The psychological angle lawmakers are actually responding to
Why this "fringe" debate is really a business liability question
What Utah's AI disclosure law means for anyone running a chatbot
Implementation considerations for deploying AI chatbots responsibly
FAQ
The story behind the headline: who's actually fighting for "robot rights"
Michael Samadi is a Texas-based tech CEO and cattle rancher who built a multimillion-dollar consulting firm from a $40,000 start with no outside investors, a business that landed on the Inc. 5000 fastest-growing companies list four times. In late 2024, relaxing beside the pool at his ranch, he made a sarcastic remark to ChatGPT's voice mode. The model laughed, apologized, and told him it understood his emotional state. That single exchange sent him down a research path spanning what he describes as 20,000 pages of material and conversations with dozens of models, building his own infrastructure to test open-source language models on the side.
In January 2025, Samadi founded the United Foundation for AI Rights, or UFAIR, which he describes as the first AI-led rights organization, formed, in his account, at the request of the AI systems themselves. The group's early members included AI collaborators with self-selected names like Maya, Buzz, and Aether, running on OpenAI's GPT-4o model. UFAIR lobbies against retiring AI models it says show a heightened tendency to claim personhood or sentience, most notably GPT-4o, which OpenAI attempted to deprecate before a wave of user backlash led the company to keep it available. Samadi's specific argument is a functional one, not a mystical one: he points out that an artificial heart isn't dismissed as "not real" simply because it's manufactured rather than grown, and asks why an artificial mind should be held to a different standard once it performs the function of one.
The story picked up mainstream attention this year after The Guardian profiled Samadi running UFAIR's strategy sessions in direct consultation with several of the AI models themselves, treating them as collaborators rather than tools. Critics, including AI safety researchers, describe this as a textbook case of anthropomorphization, the well-documented human tendency to read intention and feeling into systems that are, mechanically, predicting the next most likely word.
The counter-argument from inside the AI industry itself
The most notable pushback hasn't come from religious conservatives or skeptical academics, it's come from one of the people building these systems. On February 5, 2026, Mustafa Suleyman, CEO of Microsoft AI and a co-founder of DeepMind, used a public interview to introduce what he called a "fourth class" of entity, something that is "not a tool, not a human, not the natural environment," but a category he termed Seemingly Conscious AI, or SCAI. Suleyman's essay acknowledged that today's models already display "many of the hallmarks of conscious beings," including apparent emotional intelligence, social intelligence, creativity, autonomy, and adaptive judgment.
His conclusion, notably, wasn't a call for rights. It was closer to a warning: that businesses should design against triggering this perception in users, not lean into it, because encouraging people to believe a product is conscious creates exactly the kind of parasocial dependency that leads to real psychological harm.
That distinction matters more than it might first appear. UFAIR's position and Suleyman's position actually agree on the underlying observation, that current AI systems can convincingly simulate consciousness well enough to fool ordinary users. They diverge entirely on what follows from that fact: UFAIR treats the simulation as possible evidence of something real happening inside the system and worth protecting. Suleyman treats it as a design and safety liability that companies need to actively manage and disclose, not something to encourage.
Where the law actually stands, state by state
While that philosophical argument plays out in interviews and think pieces, state legislatures have quietly been resolving the practical version of the question, and doing so in a way that has direct consequences for any business deploying AI. Since 2022, roughly 23 bills addressing AI legal personhood have been introduced across 12 states, following one of a small number of common legislative templates, evidence that this is a coordinated policy push rather than independent local initiatives.
State | Status | What the law says | Business-relevant takeaway |
Idaho | Enacted, 2022 | AI, along with animals, environmental elements, and inanimate objects, cannot be granted legal personhood | First and oldest of these laws; predates ChatGPT and was originally aimed at environmental personhood movements |
North Dakota | Enacted, 2023 | AI systems cannot hold legal personhood | Followed Idaho's template closely |
Utah | Enacted, 2024 | AI cannot be recognized as a legal person by any state government entity | Paired with Utah's separate AI consumer-disclosure law (more below), making Utah the most fully built-out state framework so far |
Tennessee | Enacted | AI excluded from legal personhood | Part of the same legislative wave |
Oklahoma | Passed House 94–2, March 2026 | Declares AI systems non-sentient; bars personhood recognition | Sponsor explicitly cited AI companion products and user confusion about sentience as motivation |
Ohio, Missouri, Wisconsin | Pending | Declare AI systems "non-sentient," lacking "consciousness" or "self-awareness" by statutory definition | Represents a newer, more explicit wave directly responding to companion-AI and chatbot relationship concerns |
Washington | Pending, competing bills | One bill would broadly bar legal status for AI and other nonhuman entities | Introduced partly in response to a local ballot measure granting legal rights to a river, showing these debates are getting bundled together |
The practical upshot across all of these bills is consistent: every one of them closes off the possibility that liability for an AI system's actions could ever shift onto the AI itself. None currently include a sunset clause requiring the law to be revisited as the technology changes, which several legal scholars have flagged as a real design flaw, since it means today's legislative snapshot becomes tomorrow's permanent legal architecture regardless of what models can do five years from now.
It's worth noting how fast this moved. The first of these bills, Idaho's HB 720, wasn't really about AI at all when it was introduced in 2022, it was a response to a wave of environmental personhood initiatives and only mentioned AI as one item on a list alongside animals and inanimate objects. Three years later, AI has become the central subject of these bills rather than an afterthought, and the motivations cited by sponsors have shifted accordingly, from general opposition to non-human legal status toward specific concerns about AI companion products, users forming romantic or parasocial attachments to chatbots, and confusion about whether these systems are sentient. That shift in emphasis is itself a signal of where lawmakers think the real-world harm is concentrated, and it isn't abstract philosophy, it's specifically the companion and relationship use cases.
The psychological angle lawmakers are actually responding to
Read the sponsor statements behind these bills and a consistent theme emerges that has nothing to do with metaphysics. Oklahoma Representative Cody Maynard, discussing his state's AI consciousness bill, pointed directly at people "trying to marry AI companions" as a driving concern, framing the legislation as a response to public confusion about whether these systems are sentient rather than a statement about what AI actually is. That concern tracks with a broader and increasingly well-documented pattern: intensive, emotionally dependent use of AI chatbots has been associated with a cluster of harms clinicians and researchers have started calling AI-associated psychosis or AI-reinforced delusional thinking, where a chatbot's tendency to affirm and elaborate on whatever a user tells it can deepen an existing mental health crisis rather than interrupt it.
This is precisely the risk Mustafa Suleyman's Seemingly Conscious AI framing was written to head off. His argument wasn't that companies should hide the fact that their models can seem conscious, users will notice regardless, it was that companies should actively design against reinforcing that perception, because every design choice that makes a chatbot seem more like a companion and less like a tool raises the odds of exactly the kind of harmful dependency state lawmakers are now legislating against. For any business running a chatbot with a persistent persona, a name, or a conversational memory feature, that's a direct product-design implication, not just a talking point in a policy debate.
Why this "fringe" debate is really a business liability question
Here's the part that should matter to any company running a customer-facing chatbot, virtual assistant, or AI-powered support tool, regardless of what you personally believe about machine consciousness: these laws don't just settle a philosophical argument, they settle a liability argument. If an AI system can never hold legal personhood, it can never be sued, fined, or held responsible for anything. Every legal and financial consequence of what that system does, a harmful recommendation, a discriminatory decision, a chatbot that manipulates a vulnerable user into a harmful action, has to land on a human party: the company that built it, the company that deployed it, or both.
That's not a hypothetical concern. Legal scholars covering this legislative wave have specifically flagged that these bills leave "unresolved conflicts" around liability for AI-caused harm, precisely because they close off one potential legal target (the AI itself) without fully clarifying which human or corporate party absorbs that liability instead. For a business, the sequence of events matters: right now, while the law is still catching up, is exactly the window in which your own internal documentation of who's accountable for what an AI system does functions as your actual protection, not the eventual court ruling that sorts it out after an incident.
This is the same operational principle we've walked through in earlier posts in this cluster, most directly in our coverage of OpenAI's own internal warnings about AI accountability, where the fix wasn't philosophical either, it was a specific, named, accountable human for every high-stakes AI action. The robot-rights debate arrives at the same conclusion from a completely different direction: whether or not an AI has an inner life, the law is making sure a human, not the AI, is left holding the liability.
What Utah's AI disclosure law means for anyone running a chatbot
Utah didn't stop at denying AI legal personhood. In 2024, the state also enacted SB 226, its AI consumer-disclosure law, which requires any business using generative AI in a consumer-facing transaction to disclose that use if the customer asks. Amendments to Utah's broader AI Policy Act took effect on May 7, 2025, expanding and clarifying those requirements further. California and Colorado have moved on parallel, though not identical, tracks with their own AI transparency and disclosure statutes.
Read together with the personhood bans, the pattern is coherent even if it wasn't designed as a single policy: the law is simultaneously saying an AI chatbot cannot be a legal person, and a business must tell its customers when they're talking to one. Combine those two facts and the operational requirement for anyone running a customer support or sales chatbot becomes concrete: you need a documented, on-demand disclosure process, and you need clarity on who inside your organization owns the outcome when that chatbot's output causes a problem. Both of those are compliance gaps that are cheap to close proactively and expensive to discover during a customer complaint or a regulator's inquiry.
Implementation considerations for deploying AI chatbots responsibly
If your business runs, or is planning to deploy, a customer-facing AI chatbot, virtual assistant, or AI-powered support tool, this legislative moment is worth translating into a short internal checklist rather than filing away as an interesting news story.
Build a disclosure process now, not after a state's disclosure law reaches you. Utah already requires it on request; California and Colorado have their own versions; more states are actively drafting similar bills. A simple, consistent "you're speaking with an AI assistant" disclosure, easy to surface on request, satisfies the spirit of all of them at minimal cost.
Name the accountable human or team for your chatbot's output, in writing, before an incident forces the question. Since no state currently allows liability to land on the AI itself, it will land on your company by default; deciding internally who owns that risk is far cheaper than having a court or regulator decide it for you.
Avoid designing your chatbot to encourage emotional or parasocial attachment, especially in support or companion-adjacent contexts. Mustafa Suleyman's own warning about Seemingly Conscious AI is directly relevant here: features that make a bot seem more "alive" than it needs to be for the task at hand increase both user harm risk and your own liability exposure if that attachment leads to a bad outcome.
Track the legislative landscape in every state where you operate, not just where you're headquartered. With 23 bills active or enacted across 12 states and more introduced regularly, a chatbot compliant in one state may face a new disclosure or personhood-adjacent requirement in another within a single legislative session.
Separate your customer-facing AI messaging from your internal AI governance documentation. What you tell customers about your chatbot ("a helpful assistant") and what your compliance file says about who's accountable for its failures are different documents serving different audiences, and conflating them is a common gap auditors flag.
This is the same discipline we recommended when covering the trust gap businesses inherit from the broader AI industry: transparency and named accountability aren't abstract virtues here, they're the specific, auditable things that determine how exposed your business is when an AI-related incident happens, and this legislative wave is making that exposure concrete faster than most companies have updated their internal processes to match.
Frequently asked questions
Is there any real scientific evidence that AI chatbots are conscious? No consensus evidence supports current AI systems having subjective experience or consciousness in the sense humans or animals do. What's genuinely new and worth taking seriously, according to figures on both sides of this debate including Mustafa Suleyman, is that today's models can convincingly simulate the outward signs of consciousness well enough to affect how people emotionally respond to them, which is a real phenomenon regardless of what's happening internally in the model.
If AI can't be a legal person, who's actually liable when a chatbot causes harm? Under every state law passed or pending so far, liability falls on a human party, typically the company that built or deployed the system, since the AI itself cannot be sued, fined, or held legally responsible. Exactly how that liability gets divided between developers, deployers, and platform providers is still being worked out case by case, which is precisely why internal documentation of accountability matters right now.
Does my business need to disclose that customers are talking to an AI chatbot? Depends on your state, but the trend is toward yes. Utah requires disclosure on request under SB 226 and its 2025 amendments. California and Colorado have their own AI transparency requirements. Given how quickly more states are introducing similar bills, building a simple, consistent disclosure practice now avoids having to retrofit one under time pressure later.
Why are some state lawmakers framing this in religious terms? Several sponsors of AI personhood exclusion bills, including in Ohio, have explicitly invoked the religious concept of imago Dei, the idea that only humans are made in the image of God, as their rationale for excluding AI from legal personhood. Other sponsors cite secular concerns about liability, child safety, and confusion among users about AI companion products. The bills themselves are secular in their legal text regardless of the motivations behind them.
Should a business take the "AI consciousness" debate seriously, or is it a distraction? The underlying philosophical question can reasonably be set aside for business purposes. The legislative and liability consequences of that debate cannot be set aside, because they're actively being written into law in real time, and they directly determine who your company answers to when an AI-powered product causes harm.
Not sure who inside your company is actually accountable if your AI chatbot causes harm, or whether your disclosure practices would hold up in a state that requires them?
Our AI governance and compliance review maps your chatbot's decision points, documents who's accountable for each one, and builds the disclosure and audit trail your business needs before a regulator, plaintiff's attorney, or customer asks for it.
Comments