According to an investigation published in The New York Times, Anthropic—the company behind the AI ​​Claude—spent months holding private meetings with theologians and religious scholars to discuss the possibility of machines developing consciousness. The matter took on greater significance when Christopher Olah, one of the company’s co-founders, gained early access to a document from Pope Leo XIV that rejected the notion that current artificial intelligences could possess “subjective experiences.”

Olah was so concerned by this stance that he even considered canceling Anthropic’s participation in a Vatican event. Ultimately, he did attend; according to those present for the discussions, he and other company representatives sought to convince the Pope’s advisors that the possibility of artificial consciousness should be taken very seriously. During his presentation, Olah stated that Anthropic researchers had identified internal structures within the models that resemble findings from human neuroscience, as well as signs of processes that could be interpreted as introspection. Nevertheless, these observations do not prove that Claude—or any other artificial intelligence—actually possesses feelings or conscious experiences.

Anthropic itself acknowledges that it does not yet have a definitive answer to this question, and it is precisely this uncertainty that makes the episode so significant. While Hinton maintains that machines may already possess subjective experiences, one of the world’s leading companies is looking for answers even outside the field of computer science. Yet, Hinton’s concerns go further: he believes we are developing increasingly powerful systems capable of outperforming humans in a wide range of tasks, and he warns of risks—from sophisticated cyberattacks to the use of AI in the creation of biological threats.

In his view, it is not enough to simply create extremely intelligent machines; we must ensure that their goals remain aligned with human safety and well-being. Hinton has even argued that we should develop artificial intelligences that genuinely care about us, rather than merely following commands. This proposal, however, raises complex questions: how can we verify whether a machine truly cares about someone, or if it has simply learned to simulate behaviors that appear to demonstrate such concern? Furthermore, if we were to discover that certain artificial intelligences possess conscious experiences, an even more difficult question would arise: what ethical and legal responsibilities would we have toward them? For now, Hinton’s statement reflects the perspective of a highly influential scientist—not a proven discovery—but it highlights a dilemma that will become increasingly central. We are learning to build machines capable of thinking and acting with unprecedented sophistication, yet we still do not know what is required for a machine to also be able to experience the world.