Years ago, a member of my family began communicating online with someone overseas. The relationship developed slowly. Trust was built. Eventually, money became part of the conversation, and a substantial amount was lost.

The matter was reported to the FBI. As far as we could tell, nothing came of it, and the money was never recovered.

That episode happened before today’s generative artificial intelligence (AI) could compose polished messages, translate conversations instantly, manufacture convincing photographs and documents, clone a familiar voice, and help the criminal sustain believable relationships with scores of potential victims at the same time.

The con artist is an ancient character. What is new is the machine now standing beside him.

That distinction matters because AI is not merely creating new kinds of fraud. It is changing the economics of old ones. INTERPOL’s 2026 Global Financial Fraud Threat Assessment describes what is happening as the “industrialization of fraud.” The organization estimates that AI-enhanced fraud is 4.5 times more profitable than traditional methods because criminals can reach more victims while making each interaction more convincing. It also warns about capable “agentic AI” systems, able to plan and carry out sequences of tasks with limited human direction, that can automate portions of a fraud campaign.

For generations, criminals faced a choice between mass deception and personalized deception. A mass mailing could reach thousands, but it usually looked crude. A confidence man could study one victim carefully, but that required time.

AI gives criminals both: personalization at industrial scale.

The FBI’s latest numbers show that this is already a serious problem. Its 2025 Internet Crime Complaint Center report recorded 22,364 complaints containing AI-related information and more than $893 million in adjusted reported losses. The FBI says AI can create convincing synthetic profiles and personalized conversations in mass quantities, while high-quality synthetic content is becoming easier to make and harder to detect.

Among those reports were more than $30 million in losses involving AI and business email compromise, schemes in which criminals may impersonate company executives or others to redirect payments. Victims also reported more than $19 million in confidence and romance scams with a likely AI connection and more than $5 million in distress scams, including schemes using voice cloning to impersonate loved ones.

Seniors bear a disproportionate share of this damage. Complainants 60 and older filed just over 3,100 of the 22,364 AI-related complaints in 2025, yet their losses topped $352 million, well over a third of the total. Older adults are more likely to answer unfamiliar numbers, less practiced at questioning what a voice or a screen can now fabricate, and more likely to have savings within easy reach.

OpenAI’s own threat reporting provides a disturbing window into how this works. In February, the company described “Operation Date Bait,” a Cambodia-linked, semi-automated romance and task scam. Operators used AI to generate advertisements aimed at men in Indonesia interested in luxury lifestyle content. An AI chatbot posed as a flirtatious receptionist. Once contact was established, human operators working with AI generated and translated manipulative messages and steered victims toward escalating payments.

Then in July, OpenAI disclosed another Cambodia-based criminal network that used ChatGPT across investment, romance, gambling, and law enforcement impersonation schemes. The network created fake personas, translated conversations, generated promotional material, and produced images of forged passports, legal notices, investment confirmations, and gambling interfaces.

AI did not invent the confidence game. It gave the confidence game an assembly line.

And the criminal does not always need to persuade you to hand over your money. In May, Google’s Threat Intelligence Group reported the first known case of a criminal using a “zero-day” exploit, an attack built around a flaw nobody had caught yet, that Google believes was developed with AI. The flaw lets attackers slip past two-factor authentication, the extra code you type after a password, on a widely used website-management tool. They still needed a stolen password to get in, but AI did the specialized work of finding and weaponizing the flaw, work that once required a skilled human hacker. Google’s engineers caught it and closed the door before criminals could launch the mass attack they had planned.

In other words, we now face two attack surfaces: our machines and our minds.

For decades, while I worked in uniform and later as a civilian in the Pentagon, we were required every year to complete updated cybersecurity training. There was a sound reason. Adversaries changed. Techniques changed. Yesterday’s precautions might not stop tomorrow’s attack.

The cybersecurity briefing has now moved from the Pentagon conference room to the American kitchen table.

So, what should families do?

First, slow down. Criminals manufacture urgency because reflection is their enemy. The FBI now urges Americans to “take a beat” before responding to unexpected demands for money, personal information, or immediate action.

Second, verify through a separate channel. If someone claims to be your bank, hang up and call a number you independently know to be legitimate. If a child or grandchild contacts you from an unfamiliar number asking for emergency money, call that person on an established number or contact another family member. The Federal Trade Commission warns that even a familiar voice can now be cloned.

Families should also establish a private word or phrase to verify identity during an emergency, a step the FBI specifically recommends. Use multi-factor authentication on important accounts, never surrender an authentication code to someone who contacts you, and treat demands for secrecy, cryptocurrency, gift cards, unusual wire transfers, or immediate payment as warning signs.

These precautions are not paranoia. They are prudence.

Scripture has something surprisingly direct to say about this problem. Proverbs 14:15 tells us, “The simple believes everything, but the prudent gives thought to his steps.” Jesus likewise instructed his disciples to “be wise as serpents and innocent as doves” (Matthew 10:16).

Christian trust does not require credulity. We are commanded to be generous, compassionate, and truthful, but we are also responsible stewards of the people and resources God has placed in our care. A criminal who exploits compassion does not make compassion foolish. He makes discernment necessary.

Years ago, one determined person overseas managed to build enough trust with someone in my family to cause a painful financial loss.

That criminal did not have today’s artificial intelligence.

We cannot eliminate deception from a fallen world, and artificial intelligence did not put dishonesty into the human heart. It has simply given dishonesty more powerful tools.

Our answer should not be fear. It should be prudence: eyes open, safeguards in place, and the wisdom to verify before we trust.