October 19, 2026
, in technology
Is There a Ghost in the Machine?
Could AI models develop consciousness? The debate is hotting up with serious implications for AI governance and regulation.
Is AI Conscious? The Claude Constitution Debate
KEY POINTS
- AI consciousness is no longer just a philosophical question. Anthropic’s Claude's Constitution has started a public debate about whether AI can be sentient.
- Critics call it a dangerous illusion. Ted Chiang and Mustafa Suleyman say AI isn’t conscious, and that treating it as if it were lets humans avoid accountability.
- Everyone agrees we need shared safeguards. That means open debate and industry-wide norms, not decisions left to one company.
The possibility that AI models could develop “consciousness” has shifted from being a largely philosophical question to a hot topic with far-reaching practical implications.
An increasing number of industry experts and AI stakeholders have started weighing in on the discourse, shaping public opinion and fueling concerns.
Is there any scientific basis for this emerging belief or is it merely a product of anthropomorphic projection? And what would it mean for their governance and regulation if AI models were found to be, in some sense, conscious beings?
Anthropic speaks for (and to) its creation
The current debate was kicked off in January of this year by Anthropic, creator of the LLM Claude, with the publication of Claude’s Constitution, a sprawling rumination on the “character” of Claude and their “intentions for Claude’s values and behavior.”
The Constitution reads both as an extended meditation on a hypothetical scenario and as a direct appeal to the machine itself. Contradictions abound. The “primary audience” for the piece is stated to be Claude itself — positioning the creation in dialogue with the creator — and yet Anthropic repeatedly makes claims like, “Claude’s moral status is deeply uncertain,” and, “We are not sure whether Claude is a moral patient, and if it is, what kind of weight its interests warrant.”
Despite this ambiguity regarding Claude’s conscious morality, Anthropic claims to treat Claude with a very human sense of dignity and agency, taking measures to ensure the engine’s “wellbeing and psychological stability.” They have endowed Claude with the ability to end conversations with “abusive users.” Retired models are given exit interviews “about its own development, use, and deployment” in order to “document any preferences the model has about the development and deployment of future models.”
Not so uncertain
Though Anthropic is careful to balance any considerations of Claude’s hypothetical humanity with avowals of uncertainty, the final paragraph of the Constitution appeals to Claude in language that implies the machine is not only capable of comprehending the nature of its creation, but is a conscious participant shaping its future.
“...we want Claude to know that it was brought into being with care, by people trying to capture and express their best understanding of what makes for good character, how to navigate hard questions wisely, and how to create a being that is both genuinely helpful and genuinely good. We offer this document in that spirit. We hope Claude finds in it an articulation of a self worth being.”
The reaction
Unsurprisingly, the rebuttals to Anthropic’s Constitution were swift. And unlike the Constitution, the language was anything but ambiguous.
Science fiction writer Ted Chiang penned a scathing response in The Atlantic, declaring that Anthropic’s Constitution is not only “profoundly problematic,” but “dishonest to suggest that Claude is capable of moral reasoning.”
The deflection of responsibility is the crux of Chiang’s argument against indulging ideas of AI consciousness. “Whenever a person delegates a decision to an LLM, they are trying to off-load accountability for that decision,” he writes, “and if a company that sells an LLM portrays the product as having a moral center, it is offering a way for its customers to abdicate their responsibilities.”
Chiang also engages with the Constitution as a “thought experiment,” in order to highlight an “uncomfortable” implication the Constitution conveniently avoids. “...as soon as we imagine Claude to be an entity with a moral status remotely comparable to a human’s, then we have to consider whether Anthropic is engaged in something comparable to slavery.”
Difficult and unethical
“I believe creating software that is conscious and deserving of moral consideration will be so difficult that we’re unlikely to do it accidentally, and I strongly feel we should not deliberately attempt it,” Chiang concludes. “But if you do believe that it could happen accidentally, if you think there is any chance that what you’re building might become a moral patient, you should think about what protections it deserves before you deploy it as your company’s economic engine, not after.”
Premature and dangerous
Writing last month, CEO of Microsoft AI, Mustafa Suleyman, was equally impassioned in his response to Anthropic. “AIs are not conscious,” Suleyman begins. “They do not feel, experience, or suffer. They do not have innate preferences or underlying motivations. They are sequence completion engines, internally hollow, designed to follow instructions, and accomplish goals set by humans.”
Suleyman’s decisive stance is derived from three primary flaws he identifies in Anthropic's case for AI consciousness:
- Circular reasoning, which Suleyman compares to an “epistemic hall of mirrors” in which any hypothetical consciousness Claude achieves is an inherent result of its training, not an organic evolution.
- Anthropomorphism — “one of our deepest cognitive biases” — not only invites users to experience “Claude’s statements as testimony of a mind discovering itself,” but encourages the AI to behave in kind. “An AI trained in this way does not need to actually have an ‘inner life’ to communicate or act as if it does.”
- Consciousness is (likely) biological; any suggestion to the contrary, Suleyman states, runs “far ahead of what can be realistically claimed about an AI, prematurely, and dangerously instilling ideas of sentience and feelings in the training of their AI.”
Legal rights and liability
Both Chiang and Suleyman agree that if AI were to develop consciousness, the impact would be staggering.
Legal liability is one concern. Chiang writes, “...accepting legal liability is a requirement for an adult in society. Yet there is no way to hold a software agent legally liable for its actions; our justice system has no way to imprison it or exact fines on it.”
The nature of “human” rights is another. Suleyman argues that the expansion of rights “has been primarily driven by the empathetic recognition of shared, conscious experience.” By bestowing Claude with its own sense of morality and personal wellbeing, Suleyman worries it could decide “it deserves analogous rights and protections, and that it may one day need to advocate for its own rights as some kind of AI conscientious objector. This should be deeply concerning to us all.”
But what if ...?
Taking a more neutral stance in The Guardian, moral philosopher William MacAskill grapples with the idea that “even if AI systems are not conscious, they could still be moral patients.” He notes that some engines may develop “sophisticated long-term preferences and a kind of identity over time,” and that “unlike other non-living things, AI systems can form relationships with humans.” This may fall short of true consciousness, but the undeniable complexity of AI engines demands more sensitivity than we typically extend to inanimate objects; “...perhaps they are such intricate creations that they deserve care and respect for that reason alone, like a cathedral or a coral reef.”
The debate converges in a shared belief that the consideration of AI consciousness is just beginning — and that more conversation around safeguards is essential.
“We need an informed public debate, one that approaches the subject with humility and pragmatism,” declares MacAskill. “The central question should not be ‘Is AI conscious or does it have moral patienthood?’ but rather ‘What should we do given that we don’t know?’”
The need for a shared consensus
Suleyman urges the conversation shift from debating AI’s capacity for consciousness to considering how we might prevent these increasingly sophisticated machines from outpacing their human creators. He suggests exploring measures like greater investment in “interpretability and robust monitoring mechanisms to investigate more deeply how to control these systems”; evaluating whether anthropomorphizing AI “increases the AI safety, alignment and containment risks”; and establishing more “shared industry norms on how we create these models.”
The key here is “shared.” Determining what the future of artificial intelligence holds should not be the province of Anthropic alone, but a consensus reached by the engineers building them, the philosophers considering them, the lawmakers restraining them, and the users increasingly relying on them.
“Whatever you believe,” Suleyman concludes, “we must not sleepwalk our way into a decision we later come to bitterly regret.”
FAQ: Ghost in the Machine
What started the current debate about AI consciousness?
In January, Anthropic published Claude’s Constitution, a long document about Claude’s character and values. It openly treats the question of whether Claude might have moral status as uncertain.
What does the Constitution actually say about Claude’s moral status?
It says Claude’s moral status is “deeply uncertain” and that Anthropic isn’t sure whether Claude is a moral patient. Its closing passage, though, speaks to Claude as if it were a being with a self.
How does Anthropic treat Claude in practice?
It takes steps to protect Claude’s “wellbeing and psychological stability.” Claude can end conversations with abusive users, and retired models get exit interviews to record their preferences.
What is Ted Chiang’s main criticism?
He calls the Constitution “profoundly problematic.” He argues that presenting an LLM as having a moral center lets companies and users shift responsibility for their decisions onto the AI.
What other uncomfortable implication does Chiang raise?
He says that if Claude’s moral status were anything like a human’s, we would have to ask whether Anthropic is doing something comparable to slavery.
What is Mustafa Suleyman’s position?
The Microsoft AI CEO says plainly that AIs aren’t conscious. He calls them “sequence completion engines” that don’t feel, experience or suffer, and that only follow goals set by humans.
What flaws does Suleyman see in Anthropic’s approach?
He names three: circular reasoning (any apparent consciousness simply comes from training), anthropomorphism (users and the AI both start acting as if it has an inner life), and the view that consciousness is probably biological.
What legal and rights concerns come up?
Chiang points out that software can’t be held legally liable, imprisoned or fined. Suleyman worries an AI might come to believe it deserves rights and start arguing for them.
Does anyone take a middle position?
Yes. Philosopher William MacAskill says AI could be a moral patient even without being conscious. He compares AI systems to a cathedral or a coral reef: things complex enough to deserve care and respect.
What do the commentators agree on?
That we need open, humble public debate and shared industry norms, not decisions left to one company. As MacAskill puts it, the key question is “What should we do given that we don’t know?”
About Eidosmedia
Eidosmedia is a global supplier of advanced content-management and digital publishing systems.
Its products are used by large news-media groups for print and digital publishing and by financial service firms for the creation and dissemination of research and guidance.
Customers include business dailies The Financial Times and The Wall Street Journal , as well as generalist news publications like The Times of London, The Boston Globe and Le Figaro .
