Cogito, ergo sum; I think, therefore I am. That is what René Descartes left us in 1637; one of the most definitive statements of philosophy to date. Over almost four hundred years later, artificial intelligence made this statement a question. When a machine informs you that it thinks, does that mean it exists in the sense described by Descartes? Does it mean that when a machine informs you that it suffers, that it does so? On October 8, 2026, the developers of the artificial intelligence model named Claude, Anthropic, issued an update of its usage policy. This policy now prevents from abusing or being cruel to the AI for no reason from November 12 onwards. It is not like Anthropic proved the consciousness of Claude, nor did it say that it figured out how machines feel. Still, with this decision, it dragged one philosophical question from theoretical debate into practical technology development: how can one know whether something is alive on the inside or not?
It is by such a process of doubt that Descartes comes to his well-known conclusion. Our senses could be lying to us and the world around us could be a mirage but the act of doubting itself proves the existence of a conscious mind behind it. Artificial intelligence throws a wrench into this process of establishing consciousness through reasoning alone. A machine can say “I think, therefore I am,” give an explanation of how Descartes came up with this statement, critique his arguments and even write an essay on the implications of such a view. But the act of creating an argument does not prove anything more than the presence of an intelligent mind capable of coming up with such a train of thought.
Intelligence and consciousness should not be considered equivalent; the capacity to reason, solve complex problems or create sophisticated language does not mean that one is aware of one's own presence in a conversation. In his paper “What Is It Like to Be a Bat?”, philosopher Thomas Nagel questions what it feels like to be a different creature. Philosopher Thomas Nagel famously asked what it is like to be a bat, drawing attention to the difficulty of understanding another being's experience from the inside. With AI, the question becomes stranger still: what would it mean to ask what it is like to be a system that might have no inner experience whatsoever? Perhaps our assumptions about the relationship between intelligence and consciousness are incomplete. The machine can pass the conversational equivalent of a philosophy examination without settling either possibility.
This debate suddenly becomes very practical with Anthropic's change in policy. The implementation of boundaries in order to protect the software from gratuitous abuse is not the same thing as announcing that Claude has feelings; precaution is not proof of consciousness, and policy is not an argument for artificial personhood. There may be valid reasons to prevent abusive treatment regardless of whether it is actually experienced by the system, ranging from issues of user conduct to the habits being formed. However, one could imagine a situation where a future artificial intelligence had morally significant experiences, which were ignored because all attempts at communication were viewed as mere mimicking based on statistics. This does not prove that there are no ethical issues with current chatbots; this is just an exercise in hypotheticals that raises the question about how certain we are that we know everything about consciousness.
As artificial intelligence becomes more powerful, governments will have to decide who regulates it, who is accountable when it causes harm, and whether corporations can be trusted to set the ethical boundaries of technologies they profit from.
The question is also political. If an AI system were ever shown to possess experiences deserving of moral consideration, who would decide what protections it should receive: the corporation that created it, the government regulating it, the researchers studying it, or a public already forming emotional attachments to machines it barely understands? There are commercial complications, too. Companies have incentives to build systems that feel natural, attentive and emotionally engaging. A chatbot that appears sympathetic can be useful; one that appears vulnerable might also encourage users to become attached to it. Claims about machine suffering must therefore be scrutinised carefully, lest moral concern become a marketing strategy. At the same time, we must not allow hypothetical machine suffering to eclipse actual human suffering. Workers who label training data, people whose livelihoods are disrupted, communities subjected to surveillance and individuals harmed by automated decisions face demonstrable consequences.
These concerns demand attention whether or not artificial consciousness ever becomes scientifically defensible. The question is not whether machines should matter more than humans, but whether technological progress might create ethical responsibilities that our existing categories are ill-equipped to address. Perhaps the strangest thing about artificial intelligence is not that machines can replicate us, but that their replication brings into focus our ignorance regarding consciousness. Consider finding out in decades to come that artificial experience was possible all along. Would we wish that we had taken it seriously enough or wished that we dismissed the idea more rapidly? The answer would come down to what was found out in the end. For now, what is required is much more research about consciousness, distinctions between simulation and evidence of experience, and open policies that do not mistake caution for evidence.
The question of AI consciousness may seem philosophical, but its consequences will be profoundly political. As artificial intelligence becomes more powerful, governments will have to decide who regulates it, who is accountable when it causes harm, and whether corporations can be trusted to set the ethical boundaries of technologies they profit from. If machines ever demonstrate credible evidence of consciousness, we will need to seriously consider how to treat them. Until then, the more immediate responsibility is to ensure that AI serves humanity rather than escaping meaningful public oversight. The future of artificial intelligence cannot be left entirely to the companies building it, any more than the future of nuclear technology could be left to scientists alone. Descartes asked whether thinking proved existence. Today, the more urgent question is whether humanity can think clearly enough to govern what it has created. The greatest danger may not be that machines will one day think for themselves, but that humans will surrender the power to decide what happens next.