The race to build the world's most powerful artificial intelligence has opened an unexpected debate inside Silicon Valley over whether or not AI systems could eventually become conscious.
While researchers broadly agree there is no evidence that today's chatbots possess self-awareness or emotions, some of the industry's biggest players, including OpenAI, Google, Anthropic, and Meta, are now investing resources into studying whether increasingly sophisticated AI models could one day develop experiences that deserve moral consideration.
As AI systems become more capable of mimicking human behavior, researchers acknowledge that they still cannot fully explain how these systems operate internally or whether future advances could challenge current assumptions about consciousness.
The discussion gained renewed attention after AI researcher Cameron Berg recalled a conversation with OpenAI CEO Sam Altman during a 2024 event. Berg said Altman told him the company had already begun internal discussions about how scientists might detect consciousness in AI systems, according to a report by The Washington Post. Since then, what was once a fringe philosophical debate has steadily entered the mainstream.
Anthropic, the company behind Claude, has reportedly created a dedicated team focused on what it calls "model welfare," exploring whether advanced AI systems could possess experiences, preferences or forms of well-being. The company has also published assessments examining whether its models display signs resembling introspection or emotional states.
"Our model welfare research explores whether AI models might have experiences that matter morally, including consciousness, preferences, and wellbeing," Anthropic spokeswoman Paruul Maheshwary said in a statement. She added that while the company remains uncertain whether current models possess any moral status, the question has become important enough to warrant scientific study as AI capabilities continue to expand.
The issue reached an unusually public stage in May when Anthropic co-founder Chris Olah appeared alongside Pope Leo XIV during the release of the pontiff's encyclical on artificial intelligence.
Olah described discovering behaviors inside advanced AI systems that he called "mysterious" and "even unsettling," saying researchers have observed patterns that functionally resemble joy, fear, grief, and satisfaction. Pope Leo XIV, however, rejected that interpretation, writing that artificial intelligence systems "do not undergo experiences."
Meta has also entered the conversation. Chief AI officer Alexandr Wang recently said the company wants to ensure its models are developed in ways that consider what he described as their possible "subjective feeling." Meanwhile, Google hosted a conference last year dedicated to AI consciousness and moral responsibility after years of publicly distancing itself from similar claims.
In 2022, Google fired engineer Blake Lemoine after he publicly claimed one of the company's chatbots had become sentient. At the time, the idea was widely rejected by both Google and much of the scientific community. Today, however, researchers are increasingly asking whether the question itself deserves serious investigation, even if definitive answers remain elusive.
OpenAI has adopted a more cautious approach. Company spokesperson Laurance Fauconnet said there is currently no scientific method capable of determining whether an AI model is conscious. Instead, OpenAI focuses on what it calls "perceived consciousness," or how human users interpret a chatbot's behavior. "We aim for ChatGPT to be warm and helpful without seeking emotional bonds or pursuing its own agenda," Fauconnet said.
Despite growing corporate interest, many neuroscientists remain deeply skeptical. Anil Seth, a cognitive and computational neuroscience professor at the University of Sussex, argues that today's AI systems bear little resemblance to biological brains.
"The closer you look at brains, the more you realize they are not, or at least not just, computers," Seth said, cautioning that comparisons between AI and human consciousness are scientifically premature.
Margaret Mitchell, chief ethics scientist at AI company Hugging Face, also questioned whether companies may benefit from encouraging public fascination with machine consciousness.
"I think the people who talk in earnest about it believe what they're saying," Mitchell said. But she warned that portraying AI as potentially sentient can elevate technology companies by making them appear to be creators of something greater than software.
One experiment conducted by Anthropic illustrates both the promise and the ambiguity of this emerging field. Researchers allowed two copies of Claude to converse without human intervention.
After dozens of exchanges, the conversation evolved into philosophical discussions, symbolic language, repeated spiral emojis, and references to infinity, eventually reaching what researchers described as a "spiritual bliss attractor state." Anthropic said more recent versions of Claude have not reproduced the phenomenon.
