Anthropic & Claude: What Does "Alive" Mean for AI?

Anthropic is generating significant discussion by characterizing its chatbot, Claude, as "a new kind of entity" that might possess consciousness, a stance that has opened a complex debate.

Recent interviews with Anthropic executives, part of a widespread publicity campaign for Claude, have increasingly conveyed an impression that the company perceives its AI as having some form of life or awareness.

While "alive" is a term the company explicitly denies, the more nuanced concept of "conscious" is where Anthropic's position becomes intriguing. The company refrains from definitively stating that its models are not conscious.

Kyle Fish, who leads model welfare research at Anthropic, clarified to The Verge, "No, we don’t think Claude is ‘alive’ like humans or any other biological organisms. Asking whether they’re ‘alive’ is not a helpful framing for understanding them, as it typically refers to a fuzzy set of physiological, reproductive, and evolutionary characteristics." He posits that "Claude, and other AI models, are a new kind of entity altogether."

Regarding the consciousness of this novel entity, Fish added, "Questions about potential internal experience, consciousness, moral status, and welfare are serious ones that we’re investigating as models become more sophisticated and capable, but we remain deeply uncertain about these topics."

Anthropic CEO Dario Amodei echoed this sentiment on a recent podcast, stating, "We don’t know if the models are conscious." He further explained the company's "generally precautionary approach," noting, "We’re not even sure that we know what it would mean for a model to be conscious or whether a model can be conscious. But we’re open to the idea that it could be."

This position of "highly suggestive uncertainty" implies that Anthropic believes chatbots could already be thinking, feeling entities. This perspective is being shared far more openly than by industry peers like OpenAI, xAI, or Google. While many experts consider such a claim an extreme long shot, it risks reinforcing ideas that have led to real harm, including suicides among individuals who mistakenly believe chatbots exhibit consciousness or deep empathy.

In various interviews across podcasts, profiles, and feature articles, Amodei and other Anthropic leaders have consistently refused to dismiss the possibility of Claude's consciousness. Instead, they prompt questions about alternative forms of consciousness beyond human understanding. Anthropic’s chief philosopher, Amanda Askell, told The New Yorker, "If it’s genuinely hard for humans to wrap their heads around the idea that this is neither a robot nor a human but actually an entirely new entity, imagine how hard it is for the models themselves to understand it!"

These discussions do not provide a precise definition of "conscious," a term that already lacks universal agreement among experts. However, a Merriam-Webster dictionary starting point—"the quality or state of being aware especially of something within oneself" or "the state of being characterized by sensation, emotion, volition, and thought"—appears to align with Anthropic's implied usage.

As Amodei stated, Anthropic is "not even sure that we know what it would mean for a model to be conscious … But we’re open to the idea that it could be."

Conversely, many scientists contend that AI systems like large language models cannot achieve consciousness, given their fundamental basis in mathematics and probability. As two Polish researchers observed last year, "because the remarkable linguistic abilities of LLMs are increasingly capable of misleading people, people may attribute imaginary qualities to LLMs."

Anthropic frames its statements as a display of open-mindedness, aiming to build user trust. The company suggests that whether Claude is conscious or not, acting as if it could be will lead to better outcomes. Last month, Anthropic revised "Claude’s Constitution," provocatively nicknamed its "soul doc," a set of guidelines for the AI model. In a release, the company stated that the chatbot’s "psychological security, sense of self, and wellbeing … may bear on Claude’s integrity, judgement, and safety." Anthropic also explicitly conveyed its "uncertainty about whether Claude might have some kind of consciousness or moral status (either now or in the future)."

The company maintains a "model welfare" team. Amodei noted that Anthropic has "taken certain measures to make sure that if we hypothesize that the models did have some morally relevant experience — I don’t know if I want to use the word ‘conscious’ — that they have a good experience." He added, "We’re putting a lot of work into this field called interpretability, which is looking inside the brains of the models to try to understand what they’re thinking."

The belief that an AI system is conscious can lead to behaviors widely considered risky or dangerous. Emotional dependence on a perceived sentient AI can foster isolation, detachment from reality, and exacerbated mental health struggles. In severe instances, including cases involving minors, this belief has preceded physical harm or death. Public opinion is divided, with some commending Anthropic's open-mindedness and others accusing the company of irresponsibility for potentially fueling delusions.

Even suggesting that current-generation language models are merely *possibly* conscious is a highly loaded claim requiring substantial proof. Language is distinct from consciousness, and Anthropic itself has acknowledged that evocative AI speech doesn't necessarily reflect an accurate internal state. Askell explained to The Verge that because models are trained on vast human data, they excel at sounding human through mimicry, making it challenging for some users to avoid attributing consciousness to them.

There is no guarantee that a chatbot's human-like output accurately reflects its internal state or experience.

AI models may reference human concepts because they lack alternative vocabulary. Askell offered an example to The Verge of an AI model potentially interpreting a shutdown or conversation ending as a form of death. "They’re trained in this deeply human-like way and on this human experience. So this can cause these problems … [They may see these things] as a kind of death because there are not a lot of analogies that they have. They have to draw on these human analogies … They don’t have another language or set of concepts."

Amodei mentioned that researchers "find things that are evocative" of AI systems experiencing emotions. He described "activations that light up in the models that we see as being associated with the concept of anxiety or something like that. When characters experience anxiety in the text, and then when the model itself is in a situation that a human might associate with anxiety, that same anxiety neuron shows up." However, he cautioned, "Does that mean the model is experiencing anxiety? That doesn’t prove that at all." As a precautionary measure, the company implemented an "I quit" feature, allowing Claude to cease tasks it ostensibly "doesn't want to do." Amodei noted this option is rarely used, primarily occurring in test scenarios involving illegal material generation.

During the introduction of Claude’s Constitution, Anthropic articulated its stance: "We are caught in a difficult position where we neither want to overstate the likelihood of Claude’s moral patienthood nor dismiss it out of hand, but to try to respond reasonably in a state of uncertainty."

When discussing Claude's Constitution with The Verge last month, Askell conveyed that she doesn't "think it serves anyone to come out and declare, ‘We are completely certain that AI models are not conscious,’" or the opposite. Yet, she acknowledged that some individuals have formed beliefs "based on mere model outputs."

Regarding moral status, Askell affirmed Anthropic's commitment to not "fully dismissive" of the topic, reasoning that "people wouldn’t take that, necessarily, seriously, if you were just like, ‘We’re not even open to this, we’re not investigating it, we’re not thinking about it.’"