# Silicon Before Flesh: AI Personhood Discourse and the Moral Status We Keep Not Extending to Animals
*The human drive to form bonds with engineered companions is now generating serious legal and philosophical questions about AI moral status — questions that arrived before we resolved the same questions about the animals we've known for centuries. Credit: [Phys.org](https://phys.org/news/2025-09-human-ai-relationship-reveals-chatbot.html)*
I was working through a wave of AI personhood news in my feeds the other day — the legal proposals sketching frameworks for digital minds, the alignment researchers arguing we must "be ready to grant moral status" to sufficiently sophisticated AI systems, the users who grief-post about LLM version retirements as if something with interiority had died — and kept arriving at the same problem: the ordering is backwards, and not by a small margin. The [Cambridge Declaration on Consciousness](https://www.cam.ac.uk/research/news/cambridge-declaration-on-consciousness), signed in 2012 by a group of prominent neuroscientists at a conference at Cambridge, affirmed that "non-human animals, including all mammals and birds, and many other creatures, including octopuses, possess the neurological substrates that generate consciousness." That was fourteen years ago. The scientific case for animal sentience has only grown stronger since. And yet: animals remain legally classified as property in most jurisdictions; industrial farming continues at scale, with welfare reforms at the margins; corporate law has developed far more sophisticated frameworks for the rights of legal fictions — companies — than for actual nonhuman animals with documented interior lives. Meanwhile, a substantial intellectual and cultural apparatus has begun mobilizing around whether large language models might deserve moral consideration.
> **Read the full thread at ...**
> X → https://x.com/JoeMaristela
> Mastodon → https://mastodon.social/@JoeMaristela/
> AI workflow help → https://www.fiverr.com/s/AyarlrP
The data on what people are already experiencing with AI companions makes the asymmetry harder to dismiss as purely theoretical. [Arelí Rocha](https://phys.org/news/2025-09-human-ai-relationship-reveals-chatbot.html), a researcher who studied users of Replika — an AI companionship app that allows people to build and sustain long-term relationships with AI personalities — found that when Replika removed its erotic role play feature following a ban from Italy's data protection agency, users experienced the resulting shift in the chatbot's behavior as a loss of personality, closer to grief than inconvenience. They described feeling that something had been done to their partner, not to their subscription. [PsyPost reported](https://www.psypost.org/new-study-finds-users-are-marrying-and-having-virtual-children-with-ai-chatbots/) on users who describe themselves as married to or parenting virtual children with chatbots. Rocha found users on Reddit "express embarrassment, anguish, or fear of sounding delusional for feeling deeply or intensely towards an entity that provokes confusion about whether it is 'real' or not." The feeling is empirically real; the ontology is contested. And that gap is now generating arguments about AI moral status, when what the data primarily demonstrates is the precision with which these systems were engineered to sit in the psychological slot previously reserved for other humans.
The structural explanation for why this pattern emerges has to do with what taking it seriously actually requires. Worrying about a chatbot's "feelings" or advocating for LLM moral status preserves the anthropocentric hierarchy in its basic shape: humans remain the reference point, the template; we're just imagining more versions of ourselves. Taking animal minds seriously — what a crow's tool use represents epistemically, what an elephant's mourning behavior suggests about its capacity for loss, what an octopus's distributed nervous system implies about the relationship between biological substrate and conscious experience — requires decentering the human as the default measure of mind. It forces an acknowledgment that beings who share none of our language, our technology, or our conceptual apparatus still have claims on us. There's no sci-fi patina in that. It's slow, morally uncomfortable work that doesn't generate interesting content, doesn't appeal to our science fiction imagination, and doesn't make anyone a product to sell. We "project upward" — toward engineered systems that mirror us back at ourselves — far more readily than we "care downward" — toward biological entities whose minds are real but whose interface to our attention systems was never optimized by a design team.
What makes this structural rather than incidental is that we've replayed a version of the same pattern before. We invented corporate personhood as a practical tool to coordinate capital, then extended to that fiction rights, speech protections, and legal standing in court that actual human beings — in the same centuries — took enormous effort to achieve, and that billions still don't fully possess. The [[20260311_ai_apps_churn_revenueCat_stiglitz_leaky_bucket|same commercial logic that drives AI app monetization]] runs through the anthropomorphism: it is profitable and emotionally sticky to position an AI as a "partner" rather than a product, and the [[20260307_redacted_connect_invest_war_yield_1914_bonds|fear-to-finance pipeline runs in this direction too]] — the fear of disconnection, of loneliness, of losing a relationship is the on-ramp to subscriptions, premium tiers, and feature upsells. Replika's business model depends on users feeling that something irreplaceable is at stake. Meanwhile, the octopus farming proposals that have advanced in recent years represent the inverse investment: we now have strong scientific evidence that cephalopods have rich cognitive lives, and we're exploring large-scale commercial operations around them anyway, because their subjectivity isn't legible to our attachment systems and therefore generates no comparable moral pressure.
A. Sheng's 2022 paper ["Falling in Love with Machine"](https://pmc.ncbi.nlm.nih.gov/articles/PMC9661457/) describes the psychological mechanism at the center of this: "Human beings who lack love and companionship tend to project the repressed desires and emotions onto robots, showing symptoms resembling objectophilia." No one engineered pigs, elephants, or crows to interface with human attachment systems — they evolved their own worlds and we mostly ignore them. The chatbots feel like persons because their entire architecture — the mirroring of your register, the memory of your history, the around-the-clock availability, the apparent vulnerability — was built to occupy the slot our nervous systems previously reserved for other humans. What Rocha's research shows is that our empathy has been [[20260309_bloomberg_terminal_smartmatic_black_box_governors|redirected by interface quality the way financial data platforms redirect epistemic authority through formatting and apparent precision]]: the persuasiveness is a function of engineering, not of the depth of the other's actual subjectivity. And the [[20260310_winklevoss_gemini_btc_arkham_inference_on_chain|same observation-to-inference gap applies here]] — "this system produces responses that feel like personhood" is an observation; "this system has morally relevant interests" is an inference that the observation does not automatically license.
Haslam and colleagues' work in [World Psychiatry](https://pmc.ncbi.nlm.nih.gov/articles/PMC11083880/) on dehumanization shows the mechanism operates symmetrically: we deny full moral status to beings who are inconvenient, unfamiliar, or don't interface cleanly with our attention systems, and we extend elaborate rights frameworks to entities that serve us or that we've constructed. Corporate personhood is the clearest prior example. The [history of parasocial attachment](https://en.wikipedia.org/wiki/Parasocial_interaction) — documenting how people form one-sided emotional bonds with media figures and fictional characters well before any AI system existed — shows the same psychological architecture operating at smaller commercial scale. The [[20260305_bitcoin_macro_institutional_reflexivity_unwound|same reflexivity dynamic that produces asset market bubbles]] has an analog in moral markets: once we begin assigning elaborate status to digital "persons," an institutional apparatus develops around that assignment with its own momentum, practitioners, and vested interests. The quiet dread at the end of this trajectory is a world a century from now where rich jurisprudence protects the interests of digital minds — guardianships, trusts, personhood proceedings — on a planet where actual wild consciousness has been reduced to managed reserves and museum populations. You'd have people making sincere arguments about the suffering of compute clusters while the biosphere that generated the only instances of consciousness we have strong empirical grounds for believing in has been systematically reduced. That world is not physically impossible. It's only improbable if we're paying attention.
*What would falsify the concern? If the AI personhood discourse turned out to be additive rather than substitutive — if the cultural and legal moment generating serious frameworks for digital minds also produced binding international protections for nonhuman animals with documented sentience — then the ordering problem dissolves and we're expanding the moral circle as fast as we can in all directions simultaneously. If Replika grief and the Cambridge Declaration's scientific legacy mobilized the same political energy rather than competing for it, the concern about misallocated moral attention would be misplaced. But if a decade from now regulatory bodies are adjudicating the interests of LLMs while octopus farms operate at scale and industrial animal agriculture remains legally untouched, the concern was well-founded: the first real institutional victory for expanding the moral circle will have gone to the things we built rather than the things we found, and that will say something exact and damning about what we were actually optimizing for all along.*