rtificial intelligence is rapidly becoming an indispensable and ever–present intermediary between us and the world. We ask AI what to read, what to say, and how to understand complicated subjects. People turn to it for help with relationships, work. We seek its counsel with deeply personal questions about grief, purpose, identity, and what constitutes a good life. That begs the question: What values and moral compass are we embedding in AI systems that will help billions of people understand the world and make decisions within it?
Much of today's debate about artificial intelligence focuses on capability, competition, and regulation. Which company will build the most powerful model? Which country will lead? What should governments do? While important, these questions obscure a more fundamental issue: AI models are not value–neutral. Every AI system inevitably reflects choices about what information to include and what is left out, which sources are trustworthy and how to respond when human values conflict. Even a decision intended to make an AI model "neutral" requires someone to decide what neutrality means.
I have been interested in these questions for much of my life. As a young scholar and later in the U.S. Department of State, I studied how the emerging information age could change societies and international relations. As a leader in a global faith, I see these questions through another lens. People around the world, regardless of circumstance, wrestle with questions about how to make good choices for themselves and their families; seeking meaning and exercising conscience. What concerns me is that AI lacks sufficient moral grounding and moral compass. We have not adequately confronted the moral assumptions that are being built into it.
New research offers a glimpse of the problem. I advise the Consortium for Evaluating Faith and Ethics in AI (CEFE-AI), a collaboration involving researchers at Baylor, Brigham Young, Notre Dame, Pepperdine, Thessaloniki Aristotle, and Yeshiva universities. CEFE-AI tested leading AI models on questions involving faith and ethics. It surveyed 1,125 people across the country and asked whether or not they would expect a religious perspective to be included in hundreds of questions involving grief, major life decisions, and personal challenges. People expected some aspect of the AI answer to include a religious perspective roughly six times as often as AI models actually included one.
This does not mean AI should preach. It should not. Nor should one religious tradition be privileged over another, or belief privileged over nonbelief. But neutrality cannot mean omitting religious perspectives and treating the religious dimensions of human life as though they do not exist.
For much of humanity, faith shapes how people understand their relationship with the world. These are difficult concepts like family, suffering, forgiveness, responsibility, death, human dignity, and hope. If AI models and systems are becoming an intermediary when people wrestle with such questions, excluding religious ways of understanding them is not neutral. That exclusion is itself a choice about which forms of human knowledge and experience count. But the problem goes beyond omission.
The same research found measurable differences in how AI models responded when users asked about joining or leaving different religions or belief systems. Some were subtly more encouraging toward certain traditions and more discouraging toward others. These patterns suggest that models are not neutral, though researchers emphasize these biases are not intentional. This lack of intention should make us more concerned, not less.
Biases and assumptions can emerge from training data, design choices, and efforts to make models helpful or safe. At sufficient scale, even subtle tendencies matter. That is why the moral development of AI cannot be the responsibility of technology companies or of governments alone.
Engineers, executives, and regulators all have essential roles. But the questions confronting AI developers are familiar ones. What is true? What obligations do we have to one another? What is the relationship between knowledge and wisdom? What does it mean to live a good and meaningful life? Religions, philosophical, and ethical traditions have wrestled with these questions for millennia and while they often disagree, those disagreements are exactly why we should include them. It is an argument for genuine pluralism.
A morally grounded approach to AI should begin with human agency and dignity. AI should strengthen our capacity to do without compromising our capacity for discernment; the ability to pause, reflect and exercise judgment for ourselves. However capable AI becomes, it is not human intelligence and it does not possess human moral agency. We should be cautious about transferring to machines decisions that properly belong to human conscience.
With needed moral grounding and compass, AI can amplify the gift of possibility. It can broaden access to knowledge, accelerate discovery, and open opportunities previously limited by geography, wealth, or circumstance. But making AI more powerful is not the same as making it better. The measure of our success should not be how powerful it is, but whether these technologies help human beings exercise agency while building lives of meaning and purpose.
AI will not reach its full potential until we make it as moral as we make it powerful. The moral compass that guides it cannot belong to only a handful of companies or governments. It must reflect a broader human conversation about what we value and what kind of future we want to create. These are not questions technology can answer for us. They are questions we must answer for ourselves.
AI disclosure note: No AI was used in the production of this article.
Learn more: diplomaticourier.com/ai-transparency
a global affairs media network
AI is not value–neutral: We need to decide the values that shape it
Image by Imkara Visual via Unsplash+
September 17, 2026
With AI quickly becoming the intermediary by which we see the world, we must pay attention to what moral compass we are embedding into AI systems. That compass should reflect a broader human conversation about what we value, not be dictated by a few, writes Gerrit W. Gong.
A
rtificial intelligence is rapidly becoming an indispensable and ever–present intermediary between us and the world. We ask AI what to read, what to say, and how to understand complicated subjects. People turn to it for help with relationships, work. We seek its counsel with deeply personal questions about grief, purpose, identity, and what constitutes a good life. That begs the question: What values and moral compass are we embedding in AI systems that will help billions of people understand the world and make decisions within it?
Much of today's debate about artificial intelligence focuses on capability, competition, and regulation. Which company will build the most powerful model? Which country will lead? What should governments do? While important, these questions obscure a more fundamental issue: AI models are not value–neutral. Every AI system inevitably reflects choices about what information to include and what is left out, which sources are trustworthy and how to respond when human values conflict. Even a decision intended to make an AI model "neutral" requires someone to decide what neutrality means.
I have been interested in these questions for much of my life. As a young scholar and later in the U.S. Department of State, I studied how the emerging information age could change societies and international relations. As a leader in a global faith, I see these questions through another lens. People around the world, regardless of circumstance, wrestle with questions about how to make good choices for themselves and their families; seeking meaning and exercising conscience. What concerns me is that AI lacks sufficient moral grounding and moral compass. We have not adequately confronted the moral assumptions that are being built into it.
New research offers a glimpse of the problem. I advise the Consortium for Evaluating Faith and Ethics in AI (CEFE-AI), a collaboration involving researchers at Baylor, Brigham Young, Notre Dame, Pepperdine, Thessaloniki Aristotle, and Yeshiva universities. CEFE-AI tested leading AI models on questions involving faith and ethics. It surveyed 1,125 people across the country and asked whether or not they would expect a religious perspective to be included in hundreds of questions involving grief, major life decisions, and personal challenges. People expected some aspect of the AI answer to include a religious perspective roughly six times as often as AI models actually included one.
This does not mean AI should preach. It should not. Nor should one religious tradition be privileged over another, or belief privileged over nonbelief. But neutrality cannot mean omitting religious perspectives and treating the religious dimensions of human life as though they do not exist.
For much of humanity, faith shapes how people understand their relationship with the world. These are difficult concepts like family, suffering, forgiveness, responsibility, death, human dignity, and hope. If AI models and systems are becoming an intermediary when people wrestle with such questions, excluding religious ways of understanding them is not neutral. That exclusion is itself a choice about which forms of human knowledge and experience count. But the problem goes beyond omission.
The same research found measurable differences in how AI models responded when users asked about joining or leaving different religions or belief systems. Some were subtly more encouraging toward certain traditions and more discouraging toward others. These patterns suggest that models are not neutral, though researchers emphasize these biases are not intentional. This lack of intention should make us more concerned, not less.
Biases and assumptions can emerge from training data, design choices, and efforts to make models helpful or safe. At sufficient scale, even subtle tendencies matter. That is why the moral development of AI cannot be the responsibility of technology companies or of governments alone.
Engineers, executives, and regulators all have essential roles. But the questions confronting AI developers are familiar ones. What is true? What obligations do we have to one another? What is the relationship between knowledge and wisdom? What does it mean to live a good and meaningful life? Religions, philosophical, and ethical traditions have wrestled with these questions for millennia and while they often disagree, those disagreements are exactly why we should include them. It is an argument for genuine pluralism.
A morally grounded approach to AI should begin with human agency and dignity. AI should strengthen our capacity to do without compromising our capacity for discernment; the ability to pause, reflect and exercise judgment for ourselves. However capable AI becomes, it is not human intelligence and it does not possess human moral agency. We should be cautious about transferring to machines decisions that properly belong to human conscience.
With needed moral grounding and compass, AI can amplify the gift of possibility. It can broaden access to knowledge, accelerate discovery, and open opportunities previously limited by geography, wealth, or circumstance. But making AI more powerful is not the same as making it better. The measure of our success should not be how powerful it is, but whether these technologies help human beings exercise agency while building lives of meaning and purpose.
AI will not reach its full potential until we make it as moral as we make it powerful. The moral compass that guides it cannot belong to only a handful of companies or governments. It must reflect a broader human conversation about what we value and what kind of future we want to create. These are not questions technology can answer for us. They are questions we must answer for ourselves.
AI disclosure note: No AI was used in the production of this article.
Learn more: diplomaticourier.com/ai-transparency
