Ask a chatbot whether it has feelings and you’ll get a polite, careful non-answer. Ask the people who build them and you’ll get something closer to “we honestly don’t know.” AI consciousness is the question of whether an AI could actually experience anything.
It’s moved from late-night dorm debates to actual job titles at AI companies.
What it means
Consciousness, in this sense, means there’s “something it’s like” to be a thing. You’re conscious. A rock isn’t (pretty sure, anyway). The question is whether an AI system could ever land on your side of that line, and if it did, whether it would deserve any moral consideration. There’s no agreed-on test, which is most of the problem. Philosophers have argued about this for centuries with just people and animals, and they still don’t agree, so AI didn’t exactly walk into a solved question.
Why it matters
Researchers have tried to make it measurable. A 2023 report by 19 researchers listed “indicator properties” pulled from theories of human consciousness, like being able to keep an eye on its own thinking. It concluded that no current AI systems are conscious, but found no obvious technical barrier to building ones that check the boxes.
At least one lab is acting like it might matter. Anthropic hired its first full-time “model welfare” researcher in 2024 and launched a program in April 2025, saying “There’s no scientific consensus on whether current or future AI systems could be conscious.” That August, it let some Claude models end abusive conversations on their own.
Not everybody wants to go there. Microsoft AI chief Mustafa Suleyman wrote in August 2025 that “Seemingly Conscious AI is inevitable and unwelcome,” arguing the illusion itself is the danger. In late September 2026, the New York Times reported that Anthropic co-founder Chris Olah had lobbied the Pope’s advisers to take the question seriously. Sam Altman posted days later that he’s “very uncomfortable” with people giving AI models religious weight.
Simple example
A really good actor can cry on cue. The tears are real, the shaky voice is real, and you feel something watching it. Whether the actor is actually sad is a separate question, and you can’t answer it by watching harder.
With an actor, you can wait for the show to end and just ask them.
With AI, the one you’d ask is the one playing the part.

