It used to be that online deception had a tell. A strange email from a foreign prince, a spelling mistake in a bank message, a phone call from a number you didn’t recognize. You could spot the danger and delete it before it touched you. That era is over, and the CNA report “5 ways online deception works in the age of AI” shows just how dramatically the rules have changed. Today, the person on your screen can look exactly like your father, sound exactly like your daughter, and know exactly what you’re afraid of. AI has gifted scammers, manipulators, and fraudsters with tools that create not just fake messages, but fake people, fake voices, and fake realities. The technical sophistication is overwhelming, but what makes these deceptions dangerous is not the technology alone — it is the way they prey on ancient human instincts. We long to trust the face of a loved one, we crave approval from a boss, we want comfort in a lonely hour, and we seek certainty in a chaotic world. These are not weaknesses we can simply delete. They are part of being human. So as we walk through the five ways this new age of deception works, we need to keep that humanity in mind — not as a flaw to be fixed, but as the reason why these attacks succeed. Because if we can understand the emotional magic behind the machine, maybe we can begin to protect ourselves without becoming paranoid and closed-off to everyone around us.
The first form of deception the CNA report explores is the deepfake — the creation of convincingly real video and audio that put words in a mouth that never spoke them. A few years ago, this seemed like a Hollywood fantasy. Now, a frail voice can be reconstructed from ten seconds of a public video, and a face can be grafted onto another person’s body with almost perfect realism. Imagine getting a voicemail from your teenage son, his voice shaking as he explains that he crashed the car, that he’s scared, that he needs money for bail before the police take him in. Your heart drops. You call him back and he doesn’t answer, because of course he’s “in custody.” But that man on the recording? It wasn’t him. A scammer pulled a video from a birthday party posted on social media, cloned the voice, and used an AI tool to make the child say things that were never said. The terror you feel, the panic, is not an error in the scheme. The panic is the scheme. Deepfakes work by bypassing our rational brain and speaking directly to the animal part of us that reacts to a child in danger. In corporate settings, this has already caused massive losses — like the Hong Kong finance worker who saw a deepfake video of a director asking for a transfer, believed it, and moved tens of millions of dollars into the wrong account. The face was right, the voice was right, the authority was unquestionable. We used to say “I’ll believe it when I see it.” Now, seeing is precisely the thing we can no longer trust. That doesn’t mean we need to live in constant suspicion, but it does mean that shocking, urgent requests from familiar faces must be verified through a separate channel — a physical phone call, a direct look in someone’s eyes, a code word shared only between family members. Because deepfakes don’t just steal your money; they steal your ability to trust your own senses.
The second way deception has evolved is through hyper-personalized phishing, or “spear-phishing,” supercharged by AI. Traditional phishing emails were like fishing nets, cast wide and clumsily, hoping to catch anyone who overlooked bad grammar. The CNA report makes it clear that those days are fading. Today, large language models can scan someone’s public social media, their LinkedIn job details, their professional network, their recent vacation photos, even the tone of their writing, and then generate a message that feels as natural as anything a real colleague would send. It might be an email from “your boss” saying, “Hey, I can’t talk right now, but I need you to transfer this invoice before the audit — same one we spoke about yesterday.” The language is calm, familiar, and shockingly specific. It mentions your name, your project, and your little inside joke from the last team meeting. Who would question that? The human vulnerability here is not stupidity — it’s helpfulness. Most employees are trying to do their jobs, to be responsive, to not waste their manager’s time. When they see a message that looks like it comes from someone in authority, their first instinct is to accommodate, not to audit. The same mechanism works on families. Imagine an elderly grandparent receiving a message from “their granddaughter” on WhatsApp, starting with “Grandma, I need you,” and containing a reference to a childhood memory that only a family member would know. How could they resist? The AI found that memory in an old Facebook post, and now it owns the key to the grandmother’s heart. The most painful part is that we are taught to be helpful our entire lives. We are rewarded for being quick, accommodating, and trusting. AI attacks are engineered to turn those very strengths into weapons. The defense isn’t to become cold and suspicious of everyone around us. It is to create a habit of slowing down, of checking in with people outside the digital message, of asking “Would my real boss actually put this in an email?” — and then having a workplace culture where it’s safe to ask. Real relationships can handle a question. A scammer in an email cannot.
The third and perhaps most unsettling form of deception is the creation of entirely synthetic identities. In the age of AI, you no longer need to be a real person to exist online. With tools that generate endless human faces, none of which belong to anyone who has ever lived, scammers can create a perfect persona. They build a convincing Facebook or Instagram profile, complete with years of memories, photos of fake vacations, friends who are also fake, and interests carefully chosen to match the person they are trying to manipulate. These synthetic identities are used everywhere: they inflate follower counts for influencers, produce glowing product reviews, and sow political divisions. But they also strike deep into everyday commerce. CNA’s report highlights how a small business owner can see their store flooded with one-star reviews from AI-built consumers. The reviews are written in a natural, varied tone, referencing “my experience” and “the wait time” as if the person had really been there. There is no customer to resolve the problem, no refund that will make them happy, no counter-argument that will soften their anger. The reputation that took a decade to build is dismantled in a day by articles from nonexistent people. And this creates a strange kind of loneliness for real customers too. When you buy something online, you look at reviews as a way of finding honest, human community. You want to hear from someone just like you who tried it and felt relieved or disappointed. When those voices are fabricated, the entire process of making safe choices becomes unmoored. We begin to doubt every compliment and every warning. The humanizing element of this is not abstract: it’s the small business owner crying in their kitchen, watching a perfect score turn into a 3.2 overnight. It’s the consumer standing in a store, phone in hand, staring at reviews and knowing that nothing can be believed. We need new ways of establishing trust — real people vouching with their own video, local relationships, community boards, and maybe a return to small circles where you know the name of the person behind the counter. Because trust was never meant to be automated.
The fourth way AI has supercharged deception is in the realm of romance scams, sometimes called “pig butchering,” and they have become more cruel, more skilled, and more widespread than ever. Before AI, a romance scammer had to spend hours writing personalized messages, remembering details, and convincing a victim that they were a real person far away in the military or on an oil rig. That was time-consuming, which limited how many people a scammer could engage. The CNA report points out that with AI chatbots, a single criminal — or criminal network — can run love affairs with thousands of victims at the same time, in multiple languages, around the clock. The chatbot learns what the victim longs to hear. It can praise the vulnerable, reassure the anxious, and soothe the lonely. It can send deepfake photographs of a smiling man with a kind face, or a woman with soft eyes, and those faces look as real as a neighbor’s. The victim falls in love with a ghost, but the feeling is entirely real. They become attached. They share their day, their fears, their secrets. Then comes the ask: a sudden emergency at work, a trapped shipment, a medical bill, or an investment opportunity that will secure the couple’s future together. The money goes out, more and more, until there is nothing left. And the financial loss, as devastating as it is, may not be the worst part. The worst part is the shame. Victims often blame themselves, not just for sending money, but for being so lonely or credulous. They hesitate to report it, because who wants to admit they fell in love with a fiction? This is the deepest cruelty of AI-driven romance fraud: it weaponizes our need for connection against us. We all want to be seen, to be loved, to believe that someone thousands of miles away would cross oceans for us. That hope is not embarrassing. It’s human. But the scammers know it, and they mine it until there is nothing left. To fight this, we cannot shame victims. We need to normalize talking openly about online relationships, to treat them with the same caution we would a stranger in a bar, and to acknowledge that everyone — no matter how intelligent — has a place where they are vulnerable.
The fifth deception is the most systemic of all: the manipulation of our entire information ecosystem. AI can generate articles that look like journalism, photographs that look like evidence, and vast armies of bot accounts that trend hashtags and shout down dissent. The CNA report describes how social media feeds are now filled with content not because it’s true, or useful, or created by humans, but because it gets a reaction. An AI model can produce hundreds of thousands of angry comments that appear to come from real citizens, all pushing the same story, all pushing a division. One minute it’s a “scientist” explaining why a medicine is dangerous; the next minute it’s a “witness” describing a crime that never happened. The images are so realistic that they appear on news shows, law-enforcement bulletins, and family group chats. The purpose is often not to convince you of a single lie, but to exhaust you. When everything looks suspicious, when every piece of news seems unverifiable, we retreat to our corners, tired and distrustful. That is exactly the desired outcome. Online deception in the age of AI is not just about taking money. It is about weakening our shared reality, making people suspicious of each other, and creating a world where no one can agree on facts, so no one can come together to fix problems. The human way to resist this is not through a single technological tool or an app that fact-checks everything. It is to rebuild our close ties, to talk with actual neighbors, to read long-form stories, to ask hard questions, and to remember that behind every viral image and every outrage-generating post is someone trying to push you toward a reaction. We need to become better, braver readers — not only of articles, but of our own emotional responses. When we notice that a post makes us feel afraid or furious in a way that rises like a reflex, we need to pause. We need to ask who placed that feeling there. The CNA report ends with a sober truth: AI will only become more convincing, more embedded in our daily lives, and more difficult to detect. The answer is not to disconnect from the world. The answer is to connect more deeply with each other — to hold onto the kind of trust that is built on shared experience, eye contact, and the long, patient work of knowing another human being. In the age of fake faces and cloned voices, the most radical thing we can do is to be present, with each other, and with our own sense of reality.