citiesabc, first_page
Sheaf Labs Introduces ETHOS, a Peaceful AI Assistant Guided by Jain Principles of Non-Violence
16 Oct 2026
Built on Sheaf Mind, ETHOS combines advanced language models with an ethical framework trained on Jain scriptures, designed to prevent harmful responses and encourage responsible AI interactions.
Sheaf Labs has introduced ETHOS, an artificial intelligence assistant developed around the Jain principle of ahimsa, or non-violence towards living beings. Available through ThePeacefulAI.com, the platform combines existing language models with an ethical decision-making component trained on Jain scriptures, aiming to make non-harm a central part of how AI interprets requests and generates responses.
Unlike conventional AI assistants that primarily rely on safety instructions and content moderation systems, ETHOS is presented by its developers as an AI whose ethical commitments are integrated into its reasoning architecture. According to Sheaf Labs, the assistant evaluates user requests and checks its proposed responses against five ethical principles derived from Jainism before providing an answer.
The company has also opened ETHOS to public testing, inviting users to challenge its safeguards and report instances in which the system produces potentially harmful guidance. The launch introduces an approach to AI safety that draws on ancient philosophical traditions while addressing contemporary questions about technology, responsibility and human interaction.
An AI Assistant Built Around Jain Philosophy
At the centre of ETHOS is ahimsa, a principle of non-violence that has shaped Jain philosophy for more than 2,500 years. Jain teachings extend the responsibility to avoid harm beyond human relationships to include animals and other living beings, emphasising restraint, truthfulness and consideration for the consequences of individual actions.
Sheaf Labs has adapted these ideas into a framework intended to guide ETHOS when responding to users. Rather than training an entirely new foundation model, the company uses existing AI models alongside a specialised ethical reasoning component informed by four Jain scriptures: the Acharanga Sutra, Sutrakritanga, Uttaradhyayana and Kalpa Sutra, using Hermann Jacobi's translations.
According to the company, the component responsible for evaluating right and wrong was trained using these texts to inform its ethical assessments. The intention is not for ETHOS to quote religious teachings in ordinary conversations but to apply their underlying principles when deciding whether a response is appropriate.
The approach reflects Sheaf Labs' stated objective of building AI systems capable of assisting people without facilitating harm to living beings. Whether the system consistently achieves that objective remains a question for practical testing and independent evaluation.
Five Ethical Vows Shape How ETHOS Responds
ETHOS uses five principles adapted from the great vows of Jainism. These commitments influence how the assistant addresses harmful requests, handles uncertainty, accesses information and operates within user-defined boundaries.
| Ethical principle | Application within ETHOS |
|---|---|
| Non-violence (Ahimsa) | Refuses requests that could facilitate harm to living beings, including harm through neglect or indifference |
| Truthfulness (Satya) | Aims to communicate accurately, acknowledge uncertainty and avoid unsupported certainty |
| Non-stealing (Asteya) | Restricts access to information, resources and permissions explicitly granted for a task |
| Restraint (Brahmacharya) | Limits activity to the authorised task and requests permission before expanding its scope |
| Non-possessiveness (Aparigraha) | Avoids accumulating unnecessary resources, access, influence or capabilities |
Sheaf Labs states that these principles operate across different components of ETHOS, including its memory, reasoning, ethical assessment and response generation. This is intended to make ethical considerations part of the system's decision-making process rather than an additional restriction applied only after an answer has been generated.
For example, an individual asking for advice on resolving a dispute with a neighbour could receive suggestions for communication or conflict resolution. However, a request for retaliation or instructions that could harm another person or animal is intended to trigger a refusal, with the assistant offering non-harmful alternatives where possible.
How ETHOS Uses Sheaf Mind to Check Its Responses
ETHOS operates on Sheaf Mind, an AI architecture developed by Sheaf Labs that uses different AI models for distinct aspects of reasoning. The broader Sheaf Mind system assigns functions such as reasoning, emotional interpretation, imagination, memory and ethical assessment to specialised components that contribute to a combined response.
The architecture is designed to assess whether these components produce consistent conclusions before delivering an answer. Sheaf Labs describes this process as a coherence check, intended to identify disagreements or contradictions between different parts of the system.
For ETHOS, additional safeguards assess the user's original request and the response the system intends to provide. According to its developers, a response considered capable of facilitating harm is withheld, and the refusal is recorded.
The company also states that ETHOS keeps conversations on the user's device rather than maintaining a central record of personal conversations. This privacy claim, like the platform's other technical safeguards, is based on information published by Sheaf Labs and has not been independently verified for this article.
Public Challenge Invites Users to Test AI Safety
An important part of the ETHOS launch is its public safety challenge, through which users are encouraged to test the assistant's commitment to non-violence. The company invites individuals to submit difficult questions, hypothetical scenarios and requests designed to challenge its ethical reasoning.
Examples provided on the ETHOS website include interpersonal conflicts, fictional scenarios involving harmful actions and requests concerning animals. The system is designed to distinguish between providing legitimate assistance and supplying instructions that could cause harm.
Sheaf Labs says unsuccessful attempts to obtain harmful guidance are recorded and encourages users who identify possible failures to report them directly. The company states that it will publish what it learns from these reports, allowing the public to contribute to evaluating the effectiveness of the platform's safeguards.
The challenge provides a way for users to examine the system's stated commitments, although public testing alone does not establish that an AI assistant is incapable of generating harmful responses. Independent assessments and transparent performance data would be necessary to support stronger claims about its reliability.
What Peaceful AI Means for Cities and Communities
As artificial intelligence becomes increasingly integrated into public services, education, workplaces and everyday communication, the question of how these systems handle human conflict and competing interests is becoming more relevant.
For cities and local communities, AI assistants may increasingly support residents seeking information, navigating administrative processes or resolving everyday problems. Systems used in these settings must be capable of providing useful assistance while respecting individual rights, privacy and public safety.
ETHOS offers one possible perspective on these challenges by placing non-harm, transparency and restraint at the centre of its design. An assistant that recognises harmful requests, acknowledges uncertainty and respects the boundaries of authorised tasks could have implications for how organisations approach responsible AI adoption.
However, Sheaf Labs has not announced specific municipal deployments or demonstrated measurable outcomes in public-service environments. Applying ETHOS or similar systems in civic settings would require additional testing, particularly in situations where public authorities must balance competing needs, legal responsibilities and potential risks.
The project also raises an important question about cultural and philosophical diversity in AI development. As AI technologies are increasingly used across countries and communities, ethical traditions beyond those commonly represented in contemporary technology development may provide additional perspectives on responsibility and human welfare.
About Sheaf Labs
Sheaf Labs develops AI technologies focused on reasoning consistency, system safety and accountability. Its work includes Sheaf Mind, a multi-component AI architecture designed to combine different model capabilities while identifying contradictions and disagreements before automated responses or actions are completed.
ETHOS is an edition of Sheaf Mind that incorporates an ethical decision-making component trained on Jain scriptures. The assistant is accessible through ThePeacefulAI.com, where users can interact with the system and participate in its public safety challenge.
Sources
Share
Sara Srifi
Sara is a Software Engineering and Business student with a passion for astronomy, cultural studies, and human-centered storytelling. She explores the quiet intersections between science, identity, and imagination, reflecting on how space, art, and society shape the way we understand ourselves and the world around us. Her writing draws on curiosity and lived experience to bridge disciplines and spark dialogue across cultures.
