Why BIHS is taking up this question
Most AI-ethics work, including the NIST, OECD and UNESCO frameworks and the EU AI Act, is strong on risk and regulation. But it usually assumes that mind is only a product of matter. That leaves the deepest questions open: can a machine be conscious, what makes a person more than an information processor, and what should technology serve?
BIHS holds that consciousness is fundamental to reality. Its research in consciousness, metaphysics and the origins of life bears directly on these questions, so BIHS can offer the field a perspective it rarely hears.
Core principles
These principles are offered as an invitation to dialogue, not as fixed rules. Each pairs a Bhāgavata idea with its counterpart in mainstream AI ethics, and BIHS welcomes responses from scientists, technologists and other traditions.
Consciousness is primary
The self (ātmā) is distinct from matter; the knower of the field is distinct from the field (Bhagavad-gītā ch. 13).
Do not claim or imply that a model is conscious, has a soul or "understands" as a person does. AI belongs to the field, not the knower.
TransparencyDignity of every living being
Equal vision of the self in all beings (Bhagavad-gītā 5.18).
No system should treat people as mere data. Protect privacy, and care for the animals and environment that AI affects.
Fairness and human rightsTruthfulness (satya)
Truthfulness is among the divine qualities (Bhagavad-gītā 16.1–3).
Models must not deceive. Label AI-generated content, cite sources and admit uncertainty.
Honesty and provenanceNon-harm (ahiṁsā)
Non-violence is a foundational virtue (Bhagavad-gītā 16.2).
Test for harm before release. Refuse uses that injure, manipulate or exploit.
SafetyService (sevā), not exploitation
Action offered for the good of all (Bhagavad-gītā 3.9, 3.25).
Judge AI by whether it serves human and spiritual flourishing, not only profit or engagement.
BeneficenceResponsibility of the actor
The one who acts bears the result; leaders set the example (Bhagavad-gītā 3.21).
Builders and deployers stay accountable. "The algorithm did it" is never an excuse.
AccountabilityDeliberate wisdom
Reflect fully, then act (Bhagavad-gītā 18.63).
Slow down for high-stakes uses, and consult widely before deploying.
PrecautionKey questions
BIHS has the most to contribute on the first two. On the rest it joins work already under way.
- Can AI be conscious? Some researchers argue future models could deserve moral consideration. A consciousness-first view separates intelligent behavior from the presence of a self.
- What is a human being? If people come to see themselves as biological computers, their sense of meaning, responsibility and worth changes.
- Truth and deception: confident falsehoods, deepfakes and persuasive manipulation.
- Bias and equal vision: models learn the prejudices in their training data.
- Relationship and dependence: people bond with chatbots as companions or even spiritual advisors.
- AI in religious life: should AI write sermons, answer scriptural questions or simulate a teacher?
- Power: a few companies and governments control the most capable models.
- Work and livelihood: what builders owe those whose work is automated.
- Privacy and consent: people's words and images used for training without permission.
- Catastrophic risk: misuse for weapons or cyberattacks, and loss of human control.
- Environmental cost: the energy and water used to train and run large models.
Guidelines
For builders of AI models
- Document where training data comes from, and respect consent and copyright.
- Test for harmful outputs, bias and deception before release, and publish the results.
- Never design systems that claim feelings, a soul or consciousness to win trust.
- Keep a human accountable for every high-stakes decision the system informs.
- Refuse clearly harmful requests, and give users a way to report problems.
For organizations adopting AI
- Use AI to assist scholars and staff, never as the final authority on teaching or scripture.
- Disclose when content was drafted or substantially shaped by AI.
- Keep personal and donor data out of AI tools unless the provider contractually protects it.
- Review your AI policy every year.
For individuals
- Verify important claims against primary sources, especially scriptural quotations.
- Notice when time with AI replaces human relationship, study or practice.
- Treat AI as a tool, and keep your own judgment and conscience in charge.
Checklist before adopting or releasing an AI tool
Your ticks are saved only in your own browser.
Safety
AI safety risks fall into four groups, from harms happening now to longer-term ones.
| Risk | In plain words | What responsible developers do |
|---|---|---|
| Everyday errors | Confident wrong answers, bad advice, misread context | Measure accuracy, show sources, signal uncertainty, keep a human in the loop |
| Misuse | Fraud, deepfakes, harassment, cyberattacks, weapons | Usage policies, refusals, abuse monitoring, testing for dangerous capabilities |
| Societal harm | Bias, misinformation at scale, loss of privacy, manipulation, dependence | Bias audits, content labeling, privacy protection, wellbeing research |
| Loss of control | Very capable systems pursuing goals their makers did not intend | Alignment research, staged release with safety thresholds, independent government testing |
The deepest safeguard is moral: the character and intention of the people who build and direct these systems. Technical safeguards serve whatever values guide their makers.
Trusted resources
Good, free material already exists. These are the most useful starting points.
| Resource | What it offers | Cost |
|---|---|---|
| NIST AI Risk Management Framework | Practical process to map, measure and manage AI risk | Free |
| OECD AI Principles | Values adopted by 40+ countries | Free |
| UNESCO Recommendation on the Ethics of AI | Global ethics standard grounded in human dignity | Free |
| EU AI Act guide | Risk-based AI law, explained | Free |
| IEEE Ethically Aligned Design | Engineering ethics for intelligent systems | Free |
| ISO/IEC 42001 | Certifiable AI management standard | Paid |
| AI Incident Database | Real cases of AI harm, searchable | Free |
| Partnership on AI | Responsible-practice guides | Free |
| Center for AI Safety | Research and plain-language risk overviews | Free |
| Future of Life Institute | Long-term risk and policy | Free |
| UK AI Security Institute | Independent testing of advanced models | Free |
| Anthropic Responsible Scaling Policy | Example of a developer's safety thresholds | Free |
| Stanford HAI AI Index | Annual data on AI trends | Free |
| Eleos AI | Research on AI consciousness and moral status | Free |
| Ethics of AI (University of Helsinki) | Introductory online course | Free |
| BlueDot Impact | AI safety and governance courses | Free |
| Practical Data Ethics (fast.ai) | Bias, disinformation and privacy | Free |
| Montreal AI Ethics Institute | Accessible research summaries | Free |
| Rome Call for AI Ethics | Interfaith and industry pledge | Free |
| Antiqua et Nova (2025) | AI and human intelligence | Free |
| AI and Faith | Interfaith network of technologists and scholars | Free |
Study series
Six sessions introduce AI ethics through the BIHS lens. Groups, classes and families are welcome to use them.
| # | Topic | Discussion question |
|---|---|---|
| 1 | What is AI, really? | Is predicting the next word a form of understanding? |
| 2 | Consciousness and the machine | What would count as evidence that a machine is aware? |
| 3 | The human person | What can a person do that a model cannot, even in principle? |
| 4 | Truth, bias and equal vision | Whose responsibility is a biased output? |
| 5 | Safety and power | Who should decide how the most capable systems are used? |
| 6 | AI in spiritual life | Where should AI never stand in for a teacher or community? |