Kyoto University unveils Buddharoid robot monk

A research team in Kyoto has unveiled “Buddharoid,” a humanoid robot monk trained on Buddhist scriptures and designed to hold conversations while gesturing and adopting prayer postures inside temple spaces, according to Euronews citing AFP. The project, led by Seiji Kumagai at Kyoto University’s Institute for the Future of Human Society, combines a commercially available humanoid robot with large language model software.

The pitch is partly demographic. Japan’s ageing population and shrinking workforce is pushing institutions to automate tasks that used to rely on human labour, and the researchers explicitly frame the system as a way to fill gaps in religious life. In the demonstration described by AFP, the faceless, grey-robed robot moved among attendees and held one-to-one exchanges, offering answers to personal and philosophical questions.

The shift is not simply from human to machine, but from a bounded role to an unbounded interface. Earlier “religious robots” tended to be pre-programmed or limited to scripted sermons. Buddharoid is meant to be dynamic: it can improvise responses in real time, which makes it more useful as a stand-in for pastoral conversation—and harder to audit. When a priest gives bad advice, a temple can point to a person with training, supervision and a reputation to protect. When a model produces a persuasive but wrong answer, responsibility is split between the temple that deployed it, the researchers who tuned it, the vendor that built the robot body, and the model provider whose training data and guardrails are not fully visible.

That diffusion of accountability is also what makes the tool attractive. A robot monk can be available on demand, scale across locations, and offer a consistent tone. For institutions, it is a way to provide “guidance” without expanding payroll. For users, it offers the comfort of an authoritative voice without the friction of appointments or social exposure. The cost is that the content is ultimately a product of whatever data and constraints were used to train the system—and those choices are rarely disclosed in a way a layperson can evaluate.

Religious settings add another layer: the interaction is not a customer-service chat but a moral and existential consultation. The robot’s answers may be taken as tradition rather than software output, especially when delivered in a temple with ritual cues and human-like motion.

At the demonstration, the robot walked among worshippers and answered questions. The scriptures were ancient; the liability chain was new.