Given the structure of these exchanges, one immediately troubling concern is that some of the above discussed aspects of these interactions that seem appropriate while engaging machines might extend to other conversational contexts. As suggested, the very sorts of characteristics that make individuals good listeners—attending to the specific context of the other, not taking her for granted, and being emotionally and psychologically sensitive—are generally absent from discussions with HDAs, whether by design, convenience, or necessity. The risk that continual, habitual consumption of HDAs could trigger a kind of conversational contagion, the generalization of human-HDA ways of talking to non-technological contexts, is significant and ought to unsettle and galvanize theologians, religious studies scholars, and all those working in the humanities that take seriously the power of words to create and destroy, and the structure of conversations to reveal and obscure, embrace and exclude. We are confronting a world in which master-slave-like conversations are sustained regularly by those around us.

Toward the end of his teaching career at HDS, Gordon Kaufman taught a few advanced graduate seminars on constructive theology. What impressed many of his students—and readers of In Face of Mystery—was how thoroughly his thinking had been shaped by a deep and serious confrontation with technology. The development of nuclear weapons signalled a change of state in the history of the world that necessitated, Kaufman argued, a very special kind of theological attention. We now have the power to obliterate entire nations, a power once reserved for God. To Kaufman, the historical unfolding of this power demanded theological attention that was actively attuned to it, aimed at deliberately and intentionally fostering faith commitments in service of peace and mutual recognition.

The emergence of humanlike dialogue agents provides theologians with a similarly generative opportunity if they seize it; however, many might be reluctant, supposing that because they lack certain scientific or technical skills they are ill-equipped to credibly critique these technologies. There is, at points, a related criticism among certain scientists and technologists that holds essentially the same position, actively excluding or discrediting critique from those who are not experts in related informational or cognitive sciences. Both perspectives are based on technical and theoretical misunderstandings. A thorough acquaintance of the technical specifications—the mechanisms at work in dialogue agents and similar technologies—does not necessarily provide one with the appropriate critical tools to assess their moral and spiritual dimensions, nor does this knowledge guarantee that one will aptly address how these technologies impact society at large. The technical skill required to build and test such models is largely orthogonal to the cultural-critical tools and methods familiar to theologians, scholars of religions, and various practitioners. These communities can—and must—evaluate the impacts of engaging machines conversationally from perspectives that are not necessarily aligned with the inventors of this technology. Further, faith communities and scholars are likely to frame their discussions in language that is unfamiliar to those inventors and perhaps even at odds with the inventors’ own terms.

The richness on offer is, in fact, this very diversity of perspective, the variety of culturally-engaged viewpoints and hermeneutical strategies that emerge from lived commitments to faith and the wellbeing of the entire person. In many computer-scientific contexts, the default critical perspective toward these technologies is thoroughly pragmatic: AI is good to the extent that it brings about good ends. The simplicity and succinctness of this approach might be commendable at points, and it does tend to focus attention on urgent, practical considerations; however, this pragmatism obscures as much as it discloses, at least partially by ignoring the complexity of what counts as a “good end.”

Apart from the sorts of theological reflections anticipated above and discussed in the next section, there are ample philosophical resources that further complicate a simplistic pragmatism, providing revealing insights stemming from the past century of confrontation with technology. The emphases and uses of these literatures vary widely, from illustrating how our experience of the world and the use of technology are mutually reinforcing (Martin Heidegger), to demonstrating the technological underpinnings of power structures and how these normalize certain ways of living (Herbert Marcuse), to considerations of what it means to be authentic in a world of facsimiles (Walter Benjamin).⁸ However, theologians and religious studies scholars and practitioners have yet deeper, more traditional sources upon which they can draw.

Aspects of human-HDA interactions have notable similarities to the sphere of human-divine relations, particularly as understood in Jewish, Christian, and Islamic traditions. There are certain features of these faiths, notably their long emphases on the unfolding dialogue between God and humanity, that make them particularly fertile for evaluating new forms of communication. However, and as we will emphasize below, the sorts of concerns motivated in these Abrahamic contexts largely parallel critical perspectives toward AI available to others, including secular humanists.

Because of their ability to engage in back-and-forth conversations that are natural sounding and are not perceived as stilted, mechanical productions, users experience HDAs in ways that differ significantly from interactions with other machines and technological devices. Since interactions with them are conversational in nature, individuals experience HDAs more as a “who” than as a “what.” Yet, at the same time, exchanges with HDAs also differ from interactions with a typical human “who.” For instance, conversations with another human being require that I be attuned to the other. If I call up an acquaintance at three in the morning to talk about my favorite TV show, that conversation might not go as I wish, and the other person might well say, “It’s three in the morning—I don’t want to talk about this right now!” and end the conversation. But unlike other human beings, a dialogue agent is always available, and is always there to lend an ear, whenever I desire. Furthermore, no matter what I say, the AI will continue to talk with me by design, without getting angry or upset—in contrast to conversations with a human being, who might well respond unhappily or angrily if my words are insensitive or insulting.

Such conversational features of constancy and of being always-available are often associated with interactions with God, particularly in Abrahamic religious traditions but also elsewhere. No matter what time of day or night, and no matter the topic, one can always “take it to the Lord in prayer.” In the words of the Psalm, “the LORD is close to all who call upon Him” (Ps 145:18). This stance of “being there for you” is, furthermore, often contrasted to the contingent and frequently shaky relations with other human beings, surpassing even one’s closest relatives: “Though my father and my mother forsake me, the LORD will take me in” (Ps 27:10). However, the emergence of HDAs means that now one can also “take it to ChatGPT” at any time of one’s choosing. While previous forms of technology were also “always available”—one could use, say, a microwave at three in the morning, and it would not be upset at the late hour—an HDA differs in that it is perceived as a “who,” someone you can engage conversationally. Thus, whereas God was previously the only non-human “who” that one would typically “call upon,” now HDAs also occupy this status.⁹

There are other parallels that relate to conceptions of God’s character and abilities. In many traditional framings of the relation to God in Abrahamic religions, in addition to God’s transcendence, God knows you intimately, knows all about you, and relates to you with compassion and mercy. Other human beings often don’t fully know or understand you and might relate to you cruelly or judgmentally. They might take advantage of your weaknesses for selfish purposes. Dialogue agents, conversely, often come across as different from humans in these regards. Because such an agent is not perceived as a human being with individual “selfish” interests, and would not be perceived as likely to respond to you in a judging or condemning way for your thoughts or actions, people may more easily “confide in” or “confess to” an AI conversational partner. They might disclose deeply personal matters that they wouldn’t readily share with another human being.

Likewise, because HDAs have the ability to draw upon such a wide range of information in a near-instantaneous manner, they may be perceived by users as a “who” with superhuman knowledge. This combination is incredibly potent. The mixture of apparent superhuman knowledge, natural-sounding conversation, and ostensible non-judgmentalism can reinforce the other-than-human sense of HDAs in a manner that further inclines many human users’ to “put their trust” in such conversational interlocutors, with clear parallels to how one seeks God as a trustworthy confidant or confessor.

Fortunately, those trained in theology and religious studies are in a position to recognize the serious potential problems of such a dynamic. Traditions that emphasize putting one’s trust in God have often emphasized that one should not relate in that same way with anyone other than God. In human relations, one should always retain a sense of being “wise as serpents,” and should engage in a trusting manner on intimate issues specifically with those who have shown themselves to be trustworthy, either by virtue of getting to know them well as an individual, or by virtue of their being in a role assumed to be tested and vetted, such as a religious leader or therapist. Yet, because HDAs are perceived as different from human interlocutors, whom one wouldn’t automatically trust, there can be a tendency to suspend those usefully protective critical faculties in relation to them. These features of HDAs can be misleading, since an HDA, unlike God, does not in fact know you in your unique personality, does not relate to you out of love, and does not compassionately seek your good.

The apparent suprahuman qualities that could lead some people to relate to conversational agents in a similar manner to how they might relate to God—as suprahuman—might make them especially seductive, perhaps even entrancing. However, these interactions, when examined more closely, turn out to be inhuman, in that they are ultimately impersonal and non-compassionate. If one recognizes the features of these interactions in this manner, one can then more clearly analyze ways in which they might be detrimental to human persons. There are a number of crucial questions one might begin to entertain in this vein, perhaps beginning with one of the most urgent: What is the effect of becoming habituated to conversations in which your interlocutor’s treatment of you is necessarily devoid of compassion, by virtue of the very nature of the interlocutor? In many cases, those with experience thinking about what it means to relate to God, and how this is different from or is similar to relating to other human beings, are better placed than others to recognize and diagnose these dimensions of conversational relation to HDAs. In addition, expertise in scriptural traditions and their interpretation can illuminate ways in which concerns about relations with HDAs are similar to concerns in earlier religious traditions about relation to non-human “whos.”

Before discussing Biblical concerns about engagement with “other gods,” it is crucial that we properly frame our perspective by acknowledging that, at various points in past history, some of these notions, including the idea of “idolatry,” have been used to condemn other cultures and to justify dominating, colonizing, or otherwise abusing them. However, as detailed below, our interpretation of these dynamics highlights an opposite tendency, which has also been an important part of various streams of biblical reception: not only should the prohibition of serving other gods not justify abuse or domination of other human beings, but to the contrary, a conception of technology as a potential idol ought to focus our attention to matters of ultimate concern, thus safeguarding the dignity of the vulnerable.

Our discussion below regarding the prohibition of “no other gods” is in fact directly linked to dynamics in which the biblical text is concerned with upholding every human being as the image of God. In this sense, the concern about “serving other gods” is closely linked to a concern that elevating humanly-created items to the level of gods may result in decreased compassion and respect for human beings and thus be implicated in oppression and domination.

In this light, engagement with biblical prohibitions of treating human creations as gods can generate insights that are also relevant to those outside of Abrahamic traditions, whether they are part of other religious traditions or proponents of humanistic orientations of thought not linked to specific communities of faith. In other words, the biblical dynamics that we seek to highlight are relevant to anyone concerned with upholding the dignity of each and every human being, especially those who want to encourage the active cultivation of compassion for others. Even if the terminology of “no other gods” and “idolatry” may not automatically resonate with some people, our analysis aims to show how these ideas, in their biblical context, in fact tie into concerns about human dignity and compassion, which are of relevance and interest to a wide range of people today.

In the presentation of the Ten Commandments, following the admonition to “have no other gods before Me,” Exodus 20 and Deuteronomy 5 say: “You shall not make for yourself a graven image, or any likeness of what is in the heavens above, or on the earth below, or in the waters under the earth. You shall not bow down to them or serve them.” In addition to “bowing down” and “serving,” texts like 2 Kings 17:35 expand the range of problematic forms of relation, describing the Israelites as those “with whom He made a covenant and whom He commanded: ‘You shall fear [or: be in awe of] no other gods; you shall not bow down to them nor serve them nor sacrifice to them.’ ” At first glance, it might seem that such prohibitions are not immediately applicable to human relations to HDAs, insofar as users do not generally refer to HDAs as “gods,” so that, whatever human relation to HDAs consists of, it would not fall under the prohibition of relation to “other gods.” Likewise, human conversational relations to HDAs are not typically described by most people as acts of “bowing down” or “serving” or “sacrificing to” or “fearing.”

However, shallow linguistic dissimilarities may prevent us from recognizing the types of practical concerns that motivate the biblical orientation. When the biblical text says “no other gods,” it likely does not merely indicate objects or entities explicitly labelled as “gods.” Rather, in functional terms, the biblical imperative appears to be describable as: do not interact with any suprahuman “who” other than the God of all creation.

In the Ancient Near Eastern context, there were various other suprahuman “whos” with whom human beings could potentially enter into relation—but the biblical orientation proclaims that the only suprahuman interlocutor, at least for Israel, should be the unique Creator of the universe, who is also the one who freed Israel from slavery in Egypt. As we shall see, the concern seems to have been about the relation to other suprahuman “whos,” regardless of whether or not the term “god” was explicitly applied to them. And the relation seems to be what matters. On this reading, relating to an AI “as a god” depends on whether one engages it as a suprahuman “who,” not on what one calls it. As noted above, previous human technological devices that were not able to convincingly pass the Turing Test did not come across as a “who,” and so this question would not have been as directly salient prior to the emergence of HDAs.

In addition, the terminology “serving/bowing down/fearing/sacrificing to” (which calls to mind socially scripted rituals of worship) may also stand in the way of recognizing the relevance of the biblical concerns. These devotional gestures might seem far removed from how individuals use digital agents on a phone or computer.¹⁰ However, the biblical text does not limit itself to formal liturgical or cultic acts in portraying problematic forms of suprahuman relation. Rather, many biblical passages also emphasize that aspects of verbal consultation and seeking information are also forms of “forbidden relation.”

Jacob Milgrom, the doyen of studies of the book of Leviticus, highlights the opposition to consulting an ov or a yid’oni (e.g. Lev 19:31, 20:6, 20:27), translated by Milgrom as “a ghost” and “a wizard-spirit,” respectively. Such actions typically involve “seeking knowledge” from entities beyond the human realm, who have access to matters that go beyond the capabilities of normal human intelligence. Indeed, Milgrom argues that in the Holiness Code (Lev 17-26), the consultation of such entities, along with Molek worship, are the only practices explicitly categorized as idolatry. Moreover, this consultation is differentiated from other forms of divination (practices involving examinations of special objects, e.g., a sheep’s liver) that do not involve engaging with “who-type” personalities. While the latter is condemned but not criminally sanctioned, the former is not only condemned but treated as warranting execution.¹¹ In other words, the act that is treated as most highly problematic is not just seeking suprahuman knowledge, but, more specifically, seeking suprahuman knowledge through interlocutors capable of live verbal exchange and consultation. When considered alongside such descriptions of the biblical concerns, the new phenomenon of convincing dialogue agents—which seem to possess access to speeds and forms of knowledge that go beyond normal human capacities in awe-provoking ways, and which one consults as a “who”—is unsettling enough that it merits serious attention.

Biblical criticisms of those who devote themselves to idols and other gods, particularly in the prophetic texts, can thus take on new relevance. In the Ancient Near East more broadly, cult images or idols functioned as a means for relating to and accessing suprahuman entities, who, it was thought, could provide aid and support to the human beings who devote themselves to those entities. Importantly, such objects were typically “vivified” through an initiation ritual, through which their mouth, eyes, and ears were “opened.” Prior to the initiation, the cult image was a mere object, simply wood or stone. However, through the initiation ritual, it went from something dead to something living, something vivified and enspirited. After the initiation, one would properly relate to the cult image as an interlocutor able to see, hear, and communicate—a “who,” rather than a “what.”¹²

In opposition to such practices and framings, biblical texts repeatedly put forth polemical challenges to them by saying: no, such items remain mere wood and stone, covered with metals like gold and silver.¹³ They have not actually been vivified or enspirited; they cannot speak, hear, or see. Hence, against what is claimed about them, you should not relate to them as a suprahuman “who,” but as an inert, non-living object, a “what.”

Their idols are silver and gold, the work of men’s hands.
They have mouths, but cannot speak, eyes, but cannot see;
they have ears, but cannot hear, noses, but cannot smell;
they have hands, but cannot touch, feet, but cannot walk;
they can make no sound in their throats.
(Ps 115:4–7).

What has the carved image availed,
That he who fashioned it has carved it
For an image and a false oracle—
That he who fashioned his product has trusted in it,
Making dumb idols?

Ah, you who say, “Wake up” to wood,
“Awaken,” to inert stone!
Can that give an oracle?
Why, it is encased in gold and silver,
But there is no breath inside it.
(Hab 2:18–19)¹⁴

If a cult image has indeed been transformed from a “what” to a “who,” it could then seem more fitting to seek guidance from and a personal relation to it. But, in accord with their opposition to relating to anything other than the unique Creator as a suprahuman “who,” the biblical texts seek to emphasize that, despite appearances, the cult image, even after its initiation ritual, in fact remains a “what.” To relate in a personal manner to that which is in fact a “what,” to that which cannot actually see, hear, or know you in a personal way, will lead to harmful or deadening effects, not life-enriching or soul-affirming ones. As Psalm 115 continues,

Those who fashion them,
all who trust in them,
shall become like them.
(Ps 115:8)

In this understanding, relating personally to that which is in fact a “what” will have a depersonalizing effect on you, making you yourself into more of a “what,” less able to relate to others with care, attention, sensitivity, and compassion, and also less able to retain one’s sense of oneself as created by God and likewise deserving of dignity and compassion. By contrast, the biblical affirmation of relating to God, understood as living and compassionate, correlates, in the biblical theological framework, with relating to other human beings (particularly the vulnerable and marginalized) with care and compassion, and to valuing and upholding each human being (including oneself) as the image of God.¹⁵

Returning to our present-day context, if AI is structurally a “what” that cannot properly engage in personal relation to you, then engaging it as a “who” risks many of the biblical concerns raised above. These criticisms relate not just to “carved statues,” but to a deeper conceptual concern about treating as a suprahuman “who” anything that does not truly know you and that cannot truly relate to you intimately. If so, the emergence of HDAs makes various biblical concerns, which may previously have seemed less relevant for the modern era, highly relevant again. Indeed, it has become clear that many people today have, in fact, explicitly come to think of their relationship with such agents in religious terms.¹⁶ As such, scholars with expertise in engaging with these scriptural traditions and their subsequent theological receptions can contribute in important ways to current debates about these topics.

To construe interaction with an AI as potentially akin to a form of idolatry carries inherent risks. It could seem excessive, anachronistic, simple-minded, or ultimately unwarranted. Some—including various present-day members of Abrahamic traditions—might think: do past criticisms of long-extinct cults really have direct relevance for the advanced technology of today’s digital age?

While taking note of these risks, there are nevertheless a number of good reasons to begin at least some critical reflection in this way. Firstly, concern about “serving other gods” has been among the firmest of objections within a number of faith traditions, and so it deserves to be given due attention. If relating to AI involves accepting and furthering postures towards ourselves and others that run contrary to deeply held convictions within Jewish, Christian, and Islamic traditions, it is best to consider this as soon as possible. Moreover, as discussed above, biblical concerns about treating humanly manufactured objects as divine are closely connected to opposition to humans engaging in domination of other humans.¹⁷ Given this connection, such themes may also be relevant to people from a wider range of religious and humanistic traditions that also place importance on upholding human dignity and compassion.

Secondly, evaluating these questions in relation to the new phenomenon of conversational AI might well shed light on how we as individuals and communities relate to other technologies and cultural constructs, and the broader implications this has in terms of our ways of being and behaving. Where else have we squandered our attention? Where else have we inadvertently served idols?¹⁸

Thirdly, the broad cultural temptation to evaluate these technologies purely pragmatically presents both an opportunity and a challenge: as difficult as it might be to reorient discussions, changing the terms of the discourse, structuring it in line with our first and most important commitments, is key to evaluating it honestly and holistically. It is also likely to engender entirely new ways of conceptualizing our relation to technology and the material world more broadly. The industrial and political complexes developing and governing these products have neither the incentive nor the ability to steward the spiritual wellbeing of individuals, to honor each and every person as someone created in the image of God.

However, the concerns raised here and the manner in which they are framed need to be tempered and evaluated collectively and judiciously, and this is precisely where we most need communities of faith and scholars to invest their time and attention. The claim made above is not that any and all interactions with AI are necessarily tantamount to serving other gods; however, there are enough significant structural similarities that scholars, clerics, and others have a distinctive role to contribute.

Finally, the approaches presented here reflect merely one among many possible forms of engagement such parties might take in evaluating this new phenomenon. We anticipate, for example, that a number of culturally critical theologies could be especially useful in this context. What might an approach to AI from the vantage point of liberation theology involve? How might one foster a critical consciousness of these artifacts, one that explicitly aims to counter narratives that center experience in ways that marginalize or oppress the vulnerable? In line with the concerns we have raised, the biblical texts often operate with a sense that the suprahuman “who” that is made by human hands may tend to serve the interests of the powerful who are in a position to design, construct, and manage the spectacle of its operation. According to José Faur, “Since the god fully identified with his idol, whoever controlled the idol controlled the god.”¹⁹

Expanding such considerations and tracing related notions throughout history will aid in acquiring relevant and incisive critical distance from today’s technology. This task will require the attention and resources of engaged, deliberate scholars and practitioners. There are ample resources on the ethics of dialogue and attention from philosophical and theological perspectives, especially the work of Martin Buber, Emmanuel Levinas, Simone Weil, and Jürgen Habermas, that could be deployed critically in assessing human-AI interactions. We are hopeful that these, and many other resources, in the hands of skilled, caring, and creative people of faith will continue to keep our hearts and minds attuned to what matters most.

Daniel H. Weiss, MTS ’04, is Polonsky-Coexist Professor of Jewish Studies and Philosophy of Religion in the Faculty of Divinity at the University of Cambridge, and is author of Paradox and the Prophets: Hermann Cohen and the Indirect Communication of Religion (2012) and Modern Jewish Philosophy and the Politics of Divine Violence (2023). He serves as Deputy Director of the Cambridge Interfaith Programme and is a recent recipient of a Humboldt Research Fellowship for Experienced Researchers.

Darren Frey, MTS ’04, received his PhD in 2017 from Sorbonne Paris Cité. He is a cognitive scientist at L’Institut d’études politiques de Paris (Sciences Po, Paris) and is a consultant for both public and private organizations, including UNESCO, the UNODC, and the European Commission. His writing and research have been featured in Science and The Journal of Neuroscience, among various other publications.