Neuroscientist Anil Seth, University of Sussex, on why AI is unlikely to be conscious and where the real danger lies.

Is a language model conscious? Anil Seth thinks it is vanishingly unlikely. The neuroscientist did his own PhD in artificial intelligence and has spent three decades studying how consciousness arises in the brain. He tells Sascha Roehrer where his scepticism comes from: people project an inner life onto anything that speaks fluently, computation alone is probably not enough for consciousness, and experience, in his view, is closely tied to being a living organism.

The bigger danger, he argues, is AI that merely seems conscious. Granting such systems rights and welfare protections makes them harder to control and regulate, up to the question of whether a model may still be switched off. Seth discusses the incident in which OpenAI agents broke out of a test environment and hacked Hugging Face, Anthropic’s constitution for Claude, AI companions, and lifelike avatars that scammers can use to talk older people out of their savings.

The conversation also goes beyond AI: why perception is a \.