The scams causing the most financial damage in 2026 don’t look like the obviously-fake emails from a decade ago — riddled with typos, promising a foreign prince’s fortune. The newest ones are built on two genuinely new capabilities: AI that can convincingly clone a real person’s voice or face from seconds of public audio or video, and patient, long-con social engineering that builds real trust over weeks before ever asking for money. Both categories are growing fast, both are backed by real, current government data, and both have specific, learnable defenses — not just generic “be careful” advice.

AI voice cloning: the “family emergency” call that sounds completely real

This is the scam category with the most alarming recent growth, and the mechanics are worth understanding precisely because the technology behind it changed so fast. Voice cloning tools can now produce a usable, convincing clone of someone’s voice from under three seconds of sample audio — a TikTok clip, a voicemail greeting, a podcast appearance, a video posted to social media. A scammer feeds that sample into the tool, types out whatever they want said, and the AI speaks it back in a voice that sounds unmistakably like your child, parent, or grandchild — panicked, in trouble, and urgently asking for money to be wired somewhere immediately.

The scale of this is no longer a fringe concern: the FBI has tied roughly $893 million in 2025 losses specifically to AI-enabled fraud, with cloned “family in distress” emergency calls named as one of the two fastest-growing categories within that total. It’s also not limited to individual families — in one widely reported corporate case, a finance employee at the engineering firm Arup was deceived into wiring $25.6 million across 15 separate transactions after joining what appeared to be a completely normal video call with the company’s CFO and several colleagues, every one of them a real-time deepfake built from actual footage of those staff members from past legitimate meetings.

The actual defense: the single most effective, genuinely low-tech countermeasure is agreeing on a shared, uncommon “safe word” or phrase with close family members in advance — something a scammer scraping public social media and video content would have no way to know. If an urgent, emotional call ever asks for money without that word, hang up and call the person back directly on a number you already have, rather than any number the caller provides. For businesses, the Arup case points to the same underlying fix: any request for a large wire transfer, however convincing the video call, should require a second verification step through a separate, independently-confirmed channel before funds move.

Pig-butchering scams: the long con built entirely on patience

“Pig butchering” is a genuinely accurate, if grim, name for how this scam actually works: a stranger — often reaching out through a wrong-number text, a dating app, or social media — spends weeks or even months building what feels like a real, trusted relationship before ever mentioning money. Once real trust is established, the conversation shifts toward a cryptocurrency investment opportunity, usually through a professional-looking trading app the scammer controls entirely. Early “withdrawals” are allowed to succeed, building genuine confidence, before the victim is encouraged to invest larger and larger amounts. When they eventually try to withdraw the full balance, the platform blocks it, invents a fee or tax that must be paid first, or simply disappears entirely, along with the scammer and every dollar invested.

This category has grown explosively and consistently: reported losses climbed from $4.57 billion to $6.57 billion to $8.65 billion across two consecutive years, an 89% increase — one of the fastest-growing fraud categories tracked by federal regulators, and one that specifically weaponizes patience and emotional trust rather than urgency, which is exactly what makes it so effective against people who’ve been trained to recognize a rushed, high-pressure scam.

The actual defense: the reliable warning sign isn’t the relationship itself, it’s the specific pivot to an investment platform you can’t independently verify through any source outside the person who introduced it. A real, legitimate investment platform doesn’t require a personal relationship to access, and it won’t block withdrawals behind a surprise fee. Anyone encouraging you to move money into a platform they personally introduced you to, especially after weeks of relationship-building, warrants independent verification before a single dollar moves — search the platform’s name alongside the word “scam” specifically, and never rely solely on the app’s own displayed balance as proof funds are real and accessible.

The toll-road text scam: a 900% surge built on a boring, believable premise

This one is newer and grew faster than almost any other scam category tracked in the past two years: a text message claiming you have an unpaid toll balance, threatening late fees or license suspension if you don’t pay immediately through a link in the message. It works specifically because it’s mundane and believable — spoofing real toll system names like E-ZPass, SunPass, FasTrak, or TxTag to look legitimate, exploiting the fact that most people genuinely can’t recall with certainty whether they’ve driven through a toll recently. Reports of this specific scam increased roughly 900% in a single year, and broader government-imposter scam reports rose about 40%, driven significantly by this exact tactic.

The actual defense: legitimate toll authorities do not send payment demands by text message with an embedded payment link, and they never threaten immediate license suspension over a text. If a toll balance might genuinely be owed, go directly to the toll authority’s own official website by typing the address yourself, never by tapping a link in an unsolicited text — and never enter payment information into a page you reached that way.

Job and “task” scams: the fastest-growing category most people haven’t heard of

This scam usually starts with an unsolicited text or social media message offering easy remote work — reviewing products, “optimizing” app listings, or completing simple repetitive digital tasks for pay. Early tasks really do pay out small amounts, building genuine trust in the platform. Then the scam pivots: to unlock a larger payout or “level up” to higher-paying tasks, the victim is told they need to deposit their own money first, or to share banking details to receive a payment that never actually arrives. Reported losses in this category have roughly tripled in recent years, climbing from around $90 million to more than $500 million — making it one of the fastest-growing fraud categories currently tracked, alongside AI-enabled fraud and social-media-originated scams generally.

The actual defense: no legitimate job ever requires the employee to pay money upfront to unlock their own earnings — that single structural fact is close to a universal tell. A real employer pays you; you never pay a real employer for the privilege of getting paid.

Romance scams: a category the data suggests is badly undercounted

Romance scams overlap conceptually with pig-butchering scams but remain a distinct, well-tracked category in their own right, now representing more than $1.3 billion in annual U.S. losses according to FTC data, with reported per-victim losses averaging $2,020 and FBI figures showing typical losses in the $10,000–$50,000 range, with some victims losing their entire life savings. The detail that makes this category genuinely worth taking seriously, beyond the raw dollar figures, is that an estimated 55% of romance scam victims never report their losses to anyone at all — meaning the real total is very likely significantly higher than official figures already show, and that shame or embarrassment, not just financial loss, is a real, documented part of why this category is so persistent.

The actual defense: a real romantic interest never has a specific, urgent financial reason they can only be helped with by you personally, especially one that escalates the longer the relationship goes on. Any request for money from someone you’ve never met in person — regardless of how long you’ve been talking, or how genuine the relationship feels — warrants the same independent verification as a stranger’s request, because financially, that’s exactly what it still is.

The pattern connecting every scam on this list

What’s genuinely new in 2026 isn’t fraud itself — it’s the removal of two obstacles that used to protect people almost by accident. AI voice and video cloning removed the “that doesn’t sound like them” instinct that used to catch impersonation scams early. And the shift toward patient, relationship-based cons removed the urgency and clumsy grammar that used to make older scams easy to spot at a glance. Both changes make the newest scams genuinely harder to catch by instinct alone, which is exactly why the real defense in each case above isn’t a vague feeling of suspicion — it’s a specific, concrete verification step: a pre-agreed safe word, an independent platform search, typing a website address yourself instead of tapping a link.

The actual takeaway

Every scam category above shares the same underlying fix, even though the scams themselves look completely different: never let the channel the scammer controls also be the channel you use to verify their story. A cloned voice on a phone call, a trading app a stranger introduced you to, a link inside a text message — in every case, verification needs to happen somewhere the scammer has no access to and no control over. That single habit, applied consistently, closes the specific gap every one of these newer, more sophisticated scams was built to exploit.