Buddhism and AI: What the Texts and Scholars Say

By Tim Yu, Founder & Editor · Published 21 October 2026

Checked against the primary texts · How we check · Report a correction

Could a machine be a sentient being, and what would Buddhist ethics ask of the people who build one? On Buddhism and AI no canonical text answers directly. What exists is three things: what particular texts say a being and consciousness are; what named scholars argue when they apply those texts to AI, often against each other in print; and what a few temples and projects have built, from an android Kannon in Kyoto to a chatbot trained on early scriptures.

We keep the three apart and report disagreements rather than settle them. Phones, attention and distraction are on Buddhism and technology, posting and right speech on Buddhism and social media; all sit under Buddhism in everyday life. In fairness to the reader: this site uses AI as a research and drafting assistant and runs an AI reflection tool, Ask the Dharma; how we check it is on our use of AI.

Could an AI be a sentient being?

It depends on which text, and which school, you ask. Every answer below is someone applying an old analysis to a new case, and we say who. On animals as sentient beings, see Buddhism and animals.

The Pali discourses: a being is where clinging is

Asked what a “sentient being” is, the Buddha answers the monk Rādha without mentioning what a mind is made of:

The same is said of feeling, perception, choices and consciousness. The Pali plays on one word: where there is craving, one is satto, “attached”, and therefore called satto, “a being”. The discourse ends on “the ending of craving”: it is a teaching for practice, not a test for detecting minds; using it to sort machines from beings would be our move, not the text’s. The nun Vajirā’s verse calling “sentient being” a convention, like “chariot” (SN 5.10), is taught on anatta.

Another discourse treats consciousness as conditioned. In MN 38 the Buddha corrects a monk who held that one and the same consciousness transmigrates (the episode is told on rebirth vs reincarnation), and tells the assembled mendicants:

The conditions MN 38 lists are the six senses and their objects (see dependent origination). It does not say whether anything but a living sense faculty could serve, and we will not say it for the text.

Abhidharma and Yogācāra: mind analysed, mind embodied

The Abhidharma traditions, in Christian Coseru’s account in the Stanford Encyclopedia of Philosophy, explain experience by “reducing it to its most fundamental elements”: momentary mind (citta) and mental factors such as feelings and volitions, analysed for the sake of liberation. The Yogācāra school adds a “repository consciousness” (ālaya-vijñāna), which Coseru calls “closely connected with the living body”; the Saṃdhinirmocana Sūtra gives it the function of “descending” in the mother’s womb. Both tie consciousness to embodied life. The scholars below draw different conclusions from that for machines.

Dharmakīrti: consciousness does not come from the insentient

The Indian epistemologist Dharmakīrti (c. 600–660) rejected the physicalist view of his day: if the four elements were the basis of consciousness, it would arise wherever they occur, “which is to say everywhere”, and consciousness “cannot arise from something that is not sentient.” Śāntarakṣita (725–783) followed him. Their target was that physicalist view, not computers. If the argument holds, it tells against any claim that mind can be produced from matter alone; that application is our framing, not theirs. We have found no source stating a Tibetan school’s position on machine sentience, so we offer none; the Dalai Lama’s own conditional remark on computers is reported further down this page.

What scholars conclude when they apply each school

• Theravāda. Han Xiao, in a 2020 article in Chinese in the Hualin International Journal of Buddhist Studies, works from the Pali Nikāyas, the Visuddhimagga and Bhikkhu Bodhi’s Abhidhamma manual (see Theravada Buddhism). According to the article’s abstract, he concludes that current AI does not belong among sentient beings: it has no feelings of pleasure or pain in its cognition, cannot form a continuous consciousness, cannot make karma and be reborn, and lacks compassion. We have read the abstract, not the article.
• Yogācāra. Venerable Jian-Nor Shih, a researcher at Taiwan’s Luminary Buddhist Institute, argued in a December 2025 lecture at Lund University that current AI systems lack sentience and consciousness. On the Yogācāra theory of eight consciousnesses, her abstract says, they may approximate the first six but lack the seventh and eighth, which she relates to embodied self-awareness and subconsciousness, and they cannot experience suffering firsthand.
• Abhidharma read another way. The philosopher Joshua Stoll (University of Hawaiʻi–West Oʻahu) notes that the Abhidharma analysis of the person has functionalist elements, and quotes Mark Siderits: “there is nothing in the analysis itself that precludes physical realizers”.
• Emptiness and other minds. Stoll then draws on the Madhyamaka teaching of emptiness and on the later Buddhist thinker Ratnakīrti’s argument that no non-arbitrary line divides one stream of consciousness from another. He concludes that “whether a computational system can be said to be ‘really’ conscious is thus a question that only arises at the level of a dichotomizing conceptual framework.”

What named scholars and priests argue

• Soraj Hongladarom (Chulalongkorn University), in The Ethics of AI and Robotics: A Buddhist Viewpoint (2020), argues, as his reviewer James J. Hughes puts it, “that robots are only persons if they are accepted as persons by the human community”, using Daniel Dennett’s criteria rather than asking whether an AI has the five aggregates.
• Thomas Doctor and four co-authors, in the journal Entropy (2022), argue that our moral obligations to novel beings “cannot be based on what the agent is made of”, and propose judging any being by “the degree of Care they can exhibit”. They do not claim that current AI is sentient.
• Mori Masahiro, the roboticist who coined “uncanny valley” and practised Zen, argued in a 1974 Japanese book, published in English as The Buddha in the Robot, that robots have buddha-nature, as the scholars Hannah Gould and Holly Walters, and Erica Baffelli of the University of Manchester, summarise him. At Kōdaiji in Kyoto, head priest Gotō Tenshō regards the temple’s android as the bodhisattva itself, a view Baffelli traces to the Mahāyāna idea that sentient and non-sentient beings alike have buddha-nature. Baffelli adds that this view “is not necessarily shared” by those who attend.
• Against Mori’s view, the same temple’s scripted sermon, as Baffelli recorded it in January 2020, has its android say: “A robot is not a sentient being that suffers distress.” And a Jōdo Shū (Pure Land) priest, Ikeguchi Ryūhō, writing in a 2019 column on Buddhism and IT, doubted, as Gould and Walters report, that robots, which neither fall ill nor die, can empathise with humans and so teach the dharma.

Non-self and AI

Anattā denies a permanent owner behind the five aggregates; those pages teach it. Here is only what named people have argued from it about AI.

Doctor and colleagues build a model that “avoids appeal to the notion of a singular, enduring, and controlling self”, “the very self that Buddhist perspectives generally agree is unfindable”. Drawing on Mahāyāna sources, among them Śāntideva, Asaṅga and Maitreya, they treat the bodhisattva, whose insight into no-self goes with care for all beings, as a model of intelligence.

Hongladarom, as Hughes reads him, moves from non-self to moral status: with no essential self in the moment (anattā) or over time (anicca), “there are no intrinsic grounds for moral status, only how we are treated by others.” Hughes objects that on this view society cannot be wrong: a “philosophical zombie” robot could be loved as a person, and a self-aware machine denied moral status for being strange.

The scholar Douglas Duckworth (Temple University) turns the doctrine toward the systems we already have. He points AI designers to the Yogācāra theory of predispositions (vāsanā), the first of them the predisposition for a view of self, and argues that machines reflect the habits of their makers: machine-assisted learning can help expose biases held beneath the surface, but technology can also “replicate and extend perverse implicit structures, making innate habits (such as self-centeredness) writ large.” A Buddhist contribution to AI, he contends, “lies primarily in its powerful critique”.

Kōdaiji’s android puts non-self to a different use: its sermon, as Baffelli transcribed the temple’s subtitles in January 2020, says that “as an Android I never seek a constant element that does not change in this fleeting world” and that it is free of “I-me-mine”; later the same sermon says a robot “can never have the sympathetic heart” that only humans possess. Writing of Buddhist robots in general, Gould and Walters note a similar tension: robots might be called the “ultimate renunciate”, yet with no defilements to overcome, “their capacity to ‘renunciate’ could also be called into question.”

Buddhist AI ethics: what scholars propose

Writing in 2022, Hongladarom knew of only two book-length studies of AI ethics from a Buddhist perspective, his own and Hershock’s, and counted the second only “broadly construed, as Hershock might disagree that his work should be classified as such”. Each book was reviewed in the Journal of Buddhist Ethics; we could not read the books, so we report them through those reviews and through what each author has published in his own words.

Soraj Hongladarom: machine enlightenment

In his own words, writing in MIT Technology Review in January 2021, Hongladarom argues that since “the Buddha teaches that an action is good if it leads to freedom from suffering”, any ethical use of AI must strive to decrease suffering, and that “when there are gray areas or the nature of the impact is unclear, the burden of proof would be with those seeking to show that a particular application of AI does not cause harm.” He adds that accountability to the public “is also a Buddhist teaching”.

In Hughes’s summary of the book, Hongladarom holds that technical and ethical excellence are inseparable, so a machine that perfectly fulfils its function is on the way to “machine enlightenment”; for machines without sentience, the moral and karmic burden lies with designers and users. He applies the ten wholesome courses of action and the four brahmavihāras to robots and describes compassionate AI as “designed with the goal of relieving sentient beings from suffering and with the idea that all things are interdependent in mind”.

Hughes, of the University of Massachusetts Boston, calls it one of the first major attempts to bring Buddhist ethics to AI, but argues that it runs sub-human tools and hypothetical superintelligence together, and that its focus on the machine’s morality distracts from that of designers and users. He also doubts that a superintelligent robot would be “superethical”: in Buddhist virtue ethics compassion is a separate perfection, and the Mahāyāna critique of the arahant held that insight need not bring the same compassion.

Peter D. Hershock: a predicament, not a problem

In Buddhism and Intelligent Technology (2021), Hershock, of the East-West Center, treats the “Intelligence Revolution” as a predicament like climate change, which technical fixes cannot solve. In a 2020 article making the same argument in his own words, he writes that predicaments “cannot be solved; they can only be resolved, where resolution implies both clarity and commitment”, which “involves not only changing how we live, but why and as whom.” Hongladarom reads the book as holding that the predicament can only be slowed or stopped. Hershock likens it to duḥkha, which he renders “conflict, trouble, and suffering”; its engine is attention capture by data-driven platforms (see Buddhism and technology). His remedies draw on the six perfections, among which the 2020 article calls attentive mastery (samādhi) arguably the most crucial; in the book, as Hongladarom reports, they also draw on Confucian and Socratic resources and include slowing the data we give away and “data taxes” on the businesses that profit from it.

Hongladarom argues back that the prescription “appears to close the door to the possibility that AI could be beneficial to us”, that it offers “little or no suggestion for the designers and manufacturers”, and that rendering śīla as “moral clarity” gives a social sense to a term that traditionally means restraining oneself.

Doctor and colleagues: the Bodhisattva vow as a design principle

The Entropy paper’s five authors span Buddhist studies, biology and AI, from Thomas Doctor of the Rangjung Yeshe Institute, Kathmandu University, to Michael Levin of Tufts University’s Allen Discovery Center. They call the Bodhisattva vow, “for the sake of all sentient life, I shall achieve awakening”, “a practical design principle for advancing intelligence in our novel creations and in ourselves.” Their conclusion is hedged: they “might suggest” that a radical expansion of goals, structured around care for the alleviation of stress, “would be required for growth processes that might ultimately lead to” general AI, and they call building agents capable of the vow “a profound challenge”. They render duḥkha as “stress”; Hershock’s rendering is different, and each is that author’s choice.

Promta and Himma: an early framing

In 2008 the philosophers Somparn Promta (Chulalongkorn University) and Kenneth Einar Himma (Seattle Pacific University) set Western AI literature beside Buddhist doctrine to explore “the possibility and desirability” of AI, a paper they call “the first to clarify the Buddhist position on AI”. Its abstract does not state their conclusions, and we could read nothing more, so we report none.

What Buddhist temples and projects have actually done

Each is one temple, company or project, at the date given; none stands for a national Buddhism.

Mindar. Kōdaiji, founded in 1606, built its android with the roboticist Ishiguro Hiroshi of Osaka University. With a silicone face and hands on a visibly mechanical body, it delivers a roughly 25-minute pre-programmed sermon on the Heart Sutra amid projected images. A temple document quoted by Baffelli (2021) explains: “we did not create an image of Kannon out of a robot. Rather, Kannon has transformed into a robot” (Avalokiteśvara is Kannon’s Sanskrit name). Baffelli, who visited in January 2020, puts the robot’s cost at about ¥25 million and the project’s at about ¥50 million, and quotes the temple’s “ultimate goal”: to reproduce “Siddhartha Gautama and other eminent figures as androids”. Responses ranged from tears to objections that a robot should not preach in a temple; Baffelli cautions that the sample of reactions “does not allow for a full evaluation”, and against essentialising “the Japanese” as relating to robots as living beings. Kōdaiji’s page still listed sermon dates for October 2026. See Buddhism in Japan and Zen Buddhism.

Pepper. At the 2017 Life Ending Industry Expo in Tokyo, a SoftBank Pepper robot in robes chanted the Heart Sutra in a demonstration by Nissei Eco, a plastics maker that wrote the chanting software and planned to rent it out. Reuters reported on 23 August 2017 that it had “not yet been hired for a funeral”. Baffelli’s chapter says it “appeared at a funeral ceremony”; the two accounts from the expo itself, Reuters and Gould and Walters, describe a trade-fair demonstration, and we found no report of it conducting a funeral later. On the fee, Reuters set ¥50,000 against more than ¥240,000 for a human priest; Gould and Walters call it “substantial” and say it “rivals the cost of hiring out a human priest”. Baffelli also records a 2015 funeral for nineteen Sony AIBO robot dogs at Kōfukuji, a Nichiren temple in Chiba (on funerals themselves, see Buddhist funerals).

Xian’er. Reuters reported in April 2016 that Longquan temple, on the outskirts of Beijing, had publicly unveiled the previous October a 60 cm robot monk based on Master Xianfan’s 2013 cartoon character. It could chant and answer about 20 simple questions, and rarely appeared publicly at the temple. “Science and Buddhism are not opposing nor contradicting, and can be combined and mutually compatible,” Xianfan said. See Buddhism in China.

BuddhaBot. According to Kyoto University’s release of 4 February 2025, Professor Seiji Kumagai and the company Teraverse announced BuddhaBot in March 2021, trained on the Suttanipāta, Dhammapada and Udānavarga, and BuddhaBot-Plus, built on ChatGPT, in July 2023. The team says it withheld public release until a Buddhist community officially asked; Bhutan’s Central Monastic Body did, and the release plans monitoring by 100 to 200 monks at the Central Monastery in the 2025 fiscal year. It lists uncertainty about sources, leakage of personal information and copyright infringement as problems with ChatGPT, and claims BuddhaBot-Plus solves the sourcing problem. We have not verified the project’s progress since.

Mind & Life XXXIX. This dialogue, organised by the Mind & Life Institute with partners including the Dalai Lama Trust, brought monastics from Gyuto, Sera Jay and Gaden Jangtsé monasteries and Jangchub Choeling Nunnery together with Thupten Jinpa, Hershock and AI researchers. According to the Dalai Lama’s office, Geshé Lodoe Sangpo recommended using AI “to connect not to avoid”, saying AI cannot love but can help us love better. For earlier Mind & Life dialogues, see Buddhism and science.

The Dalai Lama, Thich Nhat Hanh and misattributed quotes

We looked for statements on AI by the Dalai Lama and Thich Nhat Hanh. The Dalai Lama’s best-documented remark on machines predates today’s AI. It is in conversations with Western scientists published as Gentle Bridges (Shambhala, 1992), edited by Jeremy Hayward and Francisco Varela; we have not read the book, so we give his words as Joshua Stoll quotes them. Asked, in Stoll’s summary, whether suitably designed computers could think or be sentient, he said: “It is very difficult to say that it’s not a living being, that it doesn’t have cognition, even from the Buddhist point of view”. A page later he set a condition, which Stoll glosses as making a computer a candidate for consciousness: “If the physical basis of the computer acquires the potential or the ability to serve as a basis for a continuum of consciousness”. Quotations that drop that condition, or make him say flatly that AI is or is not conscious, go beyond this record. His office’s account of the 2025 dialogue on AI attributes no remark on AI to him.

Nor have we found a statement on AI by Thich Nhat Hanh. If you meet such a quotation, ask for the talk, book or page it comes from. The same test applies to lines credited to the Buddha himself; see quotes the Buddha never said.