Helping AI to “Find Religion” Misses the Point – and Foundations – of Morality
By: Daniel Dacombe
Tech companies, especially those responsible for Large Language Model (LLM) development, are frequently in the news for the latest triumphs and horrors of AI use. You can’t scroll your social media of choice without seeing the latest headlines, and even if the future of AI is uncertain at this point many experts seems to think these new digital tools are here to stay. Recently, tech companies have been making news headlines for a different reason – they have begun to invite religious leaders to aid in the “moral development” of AI systems, all with the intention of helping these models to more accurately reflect the ethical nature of humanity. Taking steps to ensure that AI models “behave” ethically (if they can be said to behave at all) is certainly a responsible step. However, the approach of turning to religious leaders for this guidance illustrates a misunderstanding on religion’s role in human morality.
Historically, many religious leaders have maintained that morality comes from their God(s) or other spiritual sources, and are laid out in the commandments or requirements of their particular religion. What we now know is that morality predates religion, and is a result of evolution. Human beings are social creatures, and our ability to survive and thrive as groups was dependant upon how well we could take care of each other. As a result, the groups that were the best at cooperating, taking care of each other, and engaging in behaviors we now consider to be morally good were also the ones most likely to pass on their genetics and cultural norms. The foundation of much of our morality is therefore not supernatural in nature, but natural – the foundations of all the moral systems we developed were fairness, reciprocity, and most importantly human empathy.
What religions did do for moral development was to add specificity to how these foundations would play out. How do we determine what is fair? Or reciprocal? When we say “lying is wrong,” what does that mean? Put some parameters down in a sacred text and you are off to the races – no more confusion about moral choices. However, this also meant that our moral norms became “frozen” in place according to how one group of people interpreted the world and the needs of their people at that time. This is one of the reasons why as our world becomes more complex, these static moral norms in one religion or another begin to run into issues and are often reinterpreted, or elevated above new data as we discover them – which is what we usually see in fundamentalist religious groups.
In many areas of a pluralistic society, major world religions find common values to agree upon which allow them to share public space together. These values often go back to shared concepts of fairness and reciprocity. Having conversations about responsible and ethical AI models and usage should include members from these religions, as well as humanists, Indigenous peoples, and everyone who stands to be impacted from AI development. Seeing to imbue specific religions commandments, or even the foundational values those directives were based on, into AI platforms is likely to fail to produce a model that behaves morally for one single, fundamental reason: AI lacks the ability to experience empathy, which is vital to how we approach moral choices in our everyday lives.
Empathy helps us to see through the eyes of the Other, and to make choices informed by our understanding of their needs and experiences. Empathy is why people can interpret a static and rigid commandment such as “lying is wrong” in different ways, for example, “lying is wrong unless you are lying to fascists about the Jews hiding in your attic.” So, as AI lacks the ability to experience empathy, we should be concerned that any decision-making framework based on religious commandments or even basic human values could default to the coldest, most fundamentalist interpretations.
In 2025 Humanists International released the Luxemburg Declaration on Artificial Intelligence and Human Values, which begins with this principle: “AI systems have the potential to empower and assist individuals and societies to achieve more in all aspects of human life. But they must never displace human judgment, human reason, human ethics, or human responsibility for our actions. Decisions that deeply affect people’s lives must always remain in human hands.” I think that talking about imbuing values into AI models is an important conversation, but I also think – as do many Humanists the world over – that decisions about how those values are interpreted and enacted need to remain with beings that spent the last few hundred thousand years evolving the most complex empathic abilities in the world. We cannot outsource our responsibility to each other to a computer algorithm – we, human beings, must continue to take responsibility for each other and keep building towards a more equitable, empathic society.
References:
https://www.nytimes.com/2026/09/29/us/anthropic-claude-morals-ai.html