AI Deepfakes and Truth

What deepfakes actually are

The term "deepfake" originated in 2017 to describe a specific face-swapping technique. It has since broadened to cover all AI-generated synthetic media that misrepresents a real person: video, image, or audio. The defining feature is that the content is fabricated but presented (or perceived) as real.

The technology has advanced faster than most people realize. Modern deepfake tools can generate a convincing 60-second video in under 25 minutes at essentially zero cost. Voice cloning requires as little as three seconds of source audio to produce an 85% voice match. The famous Biden robocall deepfake from the 2024 New Hampshire primary cost $1 to create and took under 20 minutes. Dark-web tools that automate deepfake generation are available for as little as $20.

Three categories of deepfake account for almost all current harm:

Face-swap and synthetic video. A real person's face is mapped onto a video they were not actually in, or a fully synthetic video shows them saying or doing things they did not say or do. This is the category most people picture when they hear "deepfake." It is the category Pope Leo XIV has been personally targeted by.

Voice cloning. A person's voice is replicated from a short audio sample and used to generate fabricated speech. Voice deepfakes increased 680% year-over-year in 2024 based on analysis of over 1.2 billion calls. The technology now enables real-time voice replication, meaning a scammer can have a live phone conversation while sounding like someone the victim trusts.

Image generation and editing. AI can generate fully synthetic images of real people, edit existing photographs to put people in situations they were never in, or "nudify" clothed photos. This last application is the source of the largest documented deepfake harm category.

The technical barrier to creating deepfakes has effectively collapsed. The barriers that remain are awareness, intent, and platform policies. None of those is a strong barrier.

What deepfakes are actually doing

This is the section most Catholic and secular treatments of deepfakes get wrong. The actual harm pattern is not what most readers expect, and getting the pattern right matters for the moral analysis.

For several years, the dominant narrative about deepfakes focused on misinformation, especially in elections. The fear was that deepfakes of political candidates would distort democratic discourse, mislead voters, and undermine public trust. Research and detection systems were built around this threat model. Substantial resources were committed to defending against it.

It largely did not happen. WIRED's AI Elections Project tracked at least 78 election-related deepfakes in 2024 across multiple countries. Their effect on outcomes was minimal. Less than 1% of all fact-checked misinformation during the 2024 election cycles was AI-generated content. A May 2026 academic paper in this area concluded bluntly that the deepfake research community "prepared for the wrong threat first."

The actual deepfake harm pattern is different from the predicted one, and more disturbing.

Non-consensual intimate imagery (NCII) dominates by an enormous margin. Multiple independent surveys converge on the same number: 96-98% of deepfake content online consists of non-consensual intimate imagery. 99-100% of NCII deepfake victims are women. The UK's National Police Chief Council reported a 1,780% increase in non-consensual deepfakes from 2019 to 2024. The Internet Watch Foundation identified 1,286 illegal AI-generated child abuse videos in early 2025 alone. This is not a peripheral problem. It is what deepfake technology is overwhelmingly being used for, and the victims are overwhelmingly women and children.

Voice-clone scams targeting families and finance workers. A 2024 McAfee study found that 1 in 4 adults have experienced an AI voice scam, and 1 in 10 has been personally targeted. The typical scam: a phone call from what sounds like a family member in distress, requesting urgent money. The voice is real-time deepfake of someone whose voice was scraped from social media or a leaked recording. 77% of voice-scam victims report financial losses. In 2024, voice cloning emerged as the leading attack vector in deepfake fraud.

Corporate financial fraud. The Arup case (Hong Kong, January 2024) became the canonical example: an employee was tricked into authorizing $25.6 million in transfers via a video call in which every other participant was an AI-generated executive. The case was not unique; it was visible. CEO impersonation deepfake fraud now targets at least 400 companies per day. The average successful deepfake fraud incident costs around $500,000, with large companies sustaining higher losses.

Public-figure deception, especially religious and political figures. This is the category Pope Leo XIV has been targeted in. By September 2025, the Vatican had to officially warn about deepfake content attributed to the new pope. The harm here is not just to the targeted figure; it is to public trust in any video or audio content involving public figures. The cost of getting fooled has gotten worse; the cost of suspicion has gotten higher.

This corrected harm pattern matters for the moral analysis. The deepfake question is not primarily a "democracy in danger" question. It is a question about the safety of women and children online, about the integrity of personal communication between family members, and about the destruction of trust as a social good. These are different problems requiring different responses, and the most pastorally urgent of them is the NCII problem that almost no Catholic outlet has addressed at the level it deserves.

The Catholic claim: faces and voices are sacred

Pope Leo XIV's 2026 World Communications Day message, "Preserving Human Voices and Faces," gave the Catholic position on deepfakes its sharpest current expression. The message is short, accessible, and worth reading in full. Its core claim is dense.

"Faces and voices are sacred," the pope wrote. "God, who created us in his image and likeness, gave them to us when he called us to life through the Word he addressed to us." The sentence has theological depth that is easy to miss. It says three things at once.

First, that faces and voices are not just biological features but sacred ones. The human face and voice carry the image of God in a way that distinguishes them from any other surface of the body or any other aspect of identity. The face is where personhood becomes visible. The voice is where it becomes audible. This is a strong claim, and it is consistent across the Catholic tradition (Aquinas, the Catechism, Vatican II) even when not stated this directly.

Second, that faces and voices were given by God. They are gifts, not just possessions. This grounds the claim that we cannot rightly use someone else's face or voice without their consent. The face and voice belong to the person to whom God gave them, not to whoever has technical access to a recording.

Third, that faces and voices are linked to the Word God addressed to us. This is the deeper theological move: the human face and voice are part of how the person responds to God's creative speech. To violate someone's face or voice is to interfere with this fundamental relationship.

Pope Leo XIV continued: "By simulating human voices, faces, emotions, and relationships, the systems known as artificial intelligence not only interfere with information ecosystems, but also encroach upon the deepest level of communication, that of human relationships. The challenge, therefore, is not technological, but anthropological. Safeguarding faces and voices ultimately means safeguarding ourselves."

This is the move that distinguishes the Catholic position from most secular AI ethics on deepfakes. The secular argument tends to be about the integrity of the information ecosystem: deepfakes distort what we believe to be true, and that's bad for democracy and trust. The Catholic argument goes deeper: deepfakes violate something about what human beings are, and the violation matters whether or not anyone is deceived. A deepfake of a person can be morally wrong even if everyone knows it is fake. The wrong is at the level of how we treat persons, not just at the level of information accuracy.

The Vatican's 2025 doctrinal note Antiqua et Nova sets up this argument philosophically. The note's broader framework about AI's relationship to human dignity provides the foundation; Pope Leo XIV's communications message applies it specifically to deepfakes.

The image of God in the human face

Why does Catholic teaching treat faces as sacred? The answer comes from a long theological tradition about the human face and the doctrine of imago Dei.

The Catholic tradition holds that the human person is made in the image of God (Genesis 1:27). This image is not located in any specific feature but in the whole person. However, the face has special significance because it is where the person becomes most visible to others. The Catechism speaks of the face as a place of encounter, where one person meets another in particular ways: looking someone in the eye, recognizing a beloved face, perceiving emotion in someone else's expression. These are not minor experiences. They are how persons become present to each other.

The Christian tradition adds a further layer through the doctrine of Christ. Jesus is the visible image of the invisible God (Colossians 1:15). The face of Christ is the human face in which God is seen, and through that face, every human face becomes intelligible as bearing the divine image. Catholic art's persistent attention to the face of Christ (icons, holy face devotions, the Shroud of Turin) reflects this theological commitment.

The voice has parallel significance. God's first creative act in Genesis is speech: "God said, let there be light." The Word made flesh (John 1) gives the human voice a special place: it is through speech that persons reveal themselves, commit themselves, address each other and address God. Prayer, vows, confession, blessing — the central acts of Christian life happen through voice. To use someone's voice falsely is to invade a faculty designed for the most personal kinds of self-revelation.

None of this is parochial Catholic mysticism. Other traditions have arrived at related conclusions through different paths. Emmanuel Levinas, the twentieth-century Jewish philosopher, made the face of the other central to his entire ethical philosophy: the face is where the moral demand arises, where the other person becomes irreducibly present as a moral claim on us. Phenomenologists like Maurice Merleau-Ponty placed special weight on the face in their accounts of intersubjectivity. The Catholic position is distinctive in its theological grounding, but the claim that faces and voices have special moral status is broadly shared.

What makes deepfakes a particular moral problem from this perspective is not that they create false images (lots of things create false images). It is that they create false images specifically of faces and voices, the faculties Catholic teaching holds to be most sacred. The deepfake is not a minor offense against truth; it is an offense against the dignity of the person whose face or voice is being simulated, against the person being deceived, and against the broader social fabric in which faces and voices function as bearers of personal presence.

The specific moral problems

Several specific moral problems follow from the Catholic framing of deepfakes. Each has its own pastoral and policy implications.

The use of women's faces and bodies without consent. Given that 96-98% of deepfake content is non-consensual intimate imagery and 99-100% of victims are women, this is the largest deepfake harm by far. The Catholic teaching is sharp: this violates the dignity of the woman whose face is being used, the dignity of the user who is being formed by a fraudulent intimate experience, and the broader integrity of how human bodies and faces function in social life. The harm exists whether or not anyone is deceived; the woman whose face appears in fabricated intimate content has been violated even if every viewer knows the content is fake. Catholic teaching has long held that the use of women's bodies as objects of consumption is a violation of dignity. AI deepfakes industrialize this violation at scale.

The corruption of communication between persons who love each other. Voice-clone scams target families. The grandparent receiving a call from what sounds like their grandchild in distress, the parent receiving a call from what sounds like their child, the spouse receiving a message from what sounds like their partner — these scams work because they exploit the most basic infrastructure of human relationship. Pope Leo XIV's framing is exactly right here: the harm is not just financial; it is anthropological. Voice scams degrade something about the basic trustworthiness of voice itself.

The exploitation of children. AI-generated child sexual abuse material is one of the fastest-growing categories of deepfake harm. The UK Internet Watch Foundation identified 1,286 illegal AI-generated child abuse videos in early 2025 alone. The category includes both fabricated images of real children (used for grooming, extortion, and trafficking) and fully synthetic images that nonetheless train consumers in patterns of abuse. Catholic teaching's longstanding protection of children's dignity gives this category special moral weight, and existing legal frameworks (which were not designed for AI-generated content) are visibly inadequate.

The undermining of public figures and public trust. Pope Leo XIV's own experience as a deepfake target is the visible case, but the same pattern affects political figures, celebrities, journalists, witnesses in legal proceedings, and others whose public presence depends on the integrity of their voice and image. The harm here is not only to the targeted figures; it is to the public's ability to trust any video or audio evidence. This is what some commentators have called the "liar's dividend": the increasing plausibility of deepfakes makes it easier for genuine bad actors to dismiss real evidence as fake.

The formation of deepfake users. A subtler concern Catholic teaching highlights: the person creating or consuming deepfakes is being formed by the act. Someone who routinely creates fabricated intimate images of women, fabricated voices to extract money from elderly relatives, or fabricated political content to deceive voters is becoming a particular kind of person. The character formation is real even when the targets are anonymous or distant. Catholic Social Teaching's persistent attention to virtue and vice gives this concern its weight: practices form persons, and deepfake creation is a practice that forms persons badly.

What to actually do

Practical orientations for citizens, parents, employees, and anyone trying to navigate the deepfake age.

For yourself: develop deepfake awareness. The single most useful skill is recognizing that any video, voice message, or image could be synthetic. This does not mean reflexive disbelief; it means appropriate verification. When something seems off, check. When something is high-stakes, verify through a second channel. When a video shows someone you trust doing something out of character, ask yourself whether this is consistent with what you know about them. Pope Leo XIV's call to "pursue AI literacy" is exactly this: not technical training, but the disposition to verify rather than assume.

For parents: have the conversation about deepfakes. Children are exposed to deepfake content (and increasingly are targets) earlier than most parents realize. AI-generated images of classmates, voice-cloned scams targeting families, and synthetic content shared in peer groups are now common. The page on letting your kid use ChatGPT covers broader parental guidance; for deepfakes specifically, the conversation needs to cover: how to recognize fakes, what to do if your image is used without consent, why creating fake images of others is morally serious (not just legally risky), and how to support a friend whose image has been deepfaked.

For employees: defend against business deepfake fraud. The Arup case, the 400-companies-per-day CEO fraud rate, and the corporate deepfake landscape generally indicate that workplace deepfake defense is now a basic professional competency. Verify high-stakes requests through a separate channel. Be suspicious of urgency combined with unusual instructions. Establish family or workplace verification protocols (a code word, a particular question only the real person could answer). The Ferrari executive who blocked a deepfake call by asking for a personal verification answer is the model: deepfakes can mimic voices but not life experience.

For victims of non-consensual deepfakes: you are not alone, and there are resources. StopNCII.org provides a free service to remove non-consensual intimate imagery from major platforms. The Cyber Civil Rights Initiative offers legal and support resources. The 988 Suicide and Crisis Lifeline (call or text in the US) is available 24/7 if you are in distress. Most importantly: the moral responsibility for deepfake harm lies with the creator and distributor of the deepfake, not with the person whose image was used. The Catholic teaching is clear that the dignity of the person targeted is not diminished by the deepfake; the violation is committed against them, not by them.

For citizens: support legislation and platform accountability. The patchwork of laws around deepfakes is visibly inadequate. The EU AI Act's transparency requirements are a step forward; US state-level laws vary widely; platform self-regulation has been weak. Catholic teaching's call for institutional structures that protect human dignity applies directly here. Supporting legislation that criminalizes non-consensual intimate deepfakes, that holds platforms accountable for distribution, and that mandates provenance markings on AI-generated content is one way of living out the principle that faces and voices are sacred.

For everyone: practice what Pope Leo XIV calls "thinking for yourself." The deepest defense against deepfakes is not technical but personal. The pope's specific challenge: do not let AI become your default oracle or your default source of truth. Algorithms tell us what we want to hear. Develop the habit of asking whether the information you are encountering is reliable, where it comes from, who benefits from your believing it. This habit is what makes deepfakes less effective; it is also what makes you less susceptible to all the other ways AI is being used to shape attention and belief.

What this is really about

Deepfakes are not just a technology problem. They are a window into what AI is doing to the basic infrastructure of human life.

For most of human history, faces and voices have been reliable. You could trust that the person speaking to you was the person they appeared to be. You could trust that the photograph in front of you represented something that happened. You could trust that the voice on the phone was your child's. Deepfakes break these assumptions at scale, and they break them at the level of the most personal kinds of communication.

The Catholic response to this is not nostalgia for a pre-AI past. It is recognition that what is being eroded is something worth defending. Faces and voices have been the medium of human relationship since human relationship has existed. Their reliability is part of what makes love, friendship, trust, and community possible. Their unreliability under AI conditions is not a minor technical inconvenience; it is a wound in the fabric of social life.

Pope Leo XIV's framing brings this into focus. The challenge, he said, "is not technological, but anthropological. Safeguarding faces and voices ultimately means safeguarding ourselves." This is what is really at stake. Not just the integrity of information ecosystems, not just the safety of fraud-vulnerable groups, but the protection of what human beings are when they meet each other face to face.

The Magnifica Humanitas encyclical, which releases May 25, 2026, will almost certainly engage this theme substantially. Pope Leo XIV has been preparing the ground for it for over a year. When the encyclical drops, this page will be updated with its specific treatment of deepfakes and truth. In the meantime, the framework is already in place: faces and voices are sacred, deepfakes violate that sacredness, and Catholic teaching offers some of the sharpest resources available for thinking about why that matters.

"We need faces and voices to speak for people again," Pope Leo XIV wrote. "We need to cherish the gift of communication as the deepest truth of humanity, to which all technological innovation should also be oriented." This is the call. The technology is not going away. What we make of it depends on whether we hold onto this framework or let it slip away.