Deepfakes – Page 2 – VerifyLabsAI

Beyond “jiggle & glitch”—how deepfake crime is evolving

An AI criminal attempts to commit financial fraud but is stopped by a human using deepfake detector technology.

July 16th 2025

Remember the early deepfakes? Those grainy, often-jiggling videos with obvious lip-sync errors? Fast forward to 2025, and those “jiggle and glitch” days are long gone. Today’s deepfakes are sophisticated, convincing and the new weapon of choice for AI-driven criminals.

Deepfakes—a worldwide playground for criminals

Gone are the days when deepfakes were just about fake celebrity videos. Now, they’re precise tools for calculated fraud and deception. Here are some of the emerging categories:

Financial fraud and business-email compromise (BEC)

Imagine a video call from your CFO instructing an urgent, high-value transfer—but it’s not them. Or a voice call from your CEO authorising a payment. We’ve seen chilling real-world cases, like a Hong Kong firm losing $25 million after a deepfake video call with their “CFO” and “colleagues.” These aren’t just one-off incidents; they are highly targeted, multi-modal attacks that combine deepfaked visuals and audio with social engineering.

Identity theft and account takeover

Biometric security, once our strong shield, is now a target. Deepfakes are being used to bypass facial recognition and voice authentication systems. Criminals use stolen data to create synthetic faces and voices, then “inject” them into verification processes, fooling systems designed to keep you safe.

Romance scams and extortion

Deepfake technology adds a terrifying new dimension to emotional manipulation. Scammers create realistic “digital twins” of victims or loved ones, exploiting personal connections for financial gain or even synthetic blackmail using fabricated intimate imagery.

Political misinformation and influencing operations

Deepfakes can create fake statements from public figures, manipulate election narratives, or spread propaganda, threatening democratic processes and public discourse at scale.

Remote job interview fraud

A new frontier of deepfake crime involves using synthetic video and audio to impersonate candidates in remote interviews, gaining access to sensitive company information or even employment under false pretenses.

Vigilance is no longer enough

The speed and accessibility of generative AI tools mean these sophisticated attacks are no longer reserved for highly skilled hackers. Off-the-shelf tools make it easier for anyone to create convincing fakes.

What does this mean for you?

Seeing isn’t believing: your eyes and ears can be fooled.
Biometrics aren’t foolproof: even advanced security systems can be bypassed.
The threat is personal: your finances, your identity and your relationships can be targets no matter who you are or what you do.

In this rapidly evolving landscape, simple vigilance and common sense, while important, are often no match for an AI-powered adversary.

It’s time to equip yourself with the proactive defenses required for the digital age.

What AI perceives—how machines unmask deepfakes

July 16th 2025

We’ve all heard the warnings about deepfakes – hyper-realistic fake images, videos, and audio created by AI. The scary truth? They’re often too good for the human eye to detect. Our brains are wired to quickly process faces and familiar patterns, but AI-generated fakes are specifically designed to fool those very systems.

So, if our eyes can’t catch them, what does? The answer lies in how AI sees and thinks differently than we do.

It’s not about “looking fake”, it’s about “being imperfect”

Imagine you’re inspecting a counterfeit banknote. You might look for obvious errors. But a machine inspects it for subtle anomalies, ink patterns, and micro-text that a human would never notice. That’s how AI approaches deepfake detection.

Instead of seeing a whole, recognisable face, deepfake detection AI processes content at a granular level, looking for microscopic inconsistencies and deviations from real-world physics and human biology.

Here’s a glimpse into what AI “sees”:

The science of light & shadow: AI understands how light behaves. It looks for even the tiniest discrepancies in how light falls on a face versus the background, or shadows that don’t quite match the presumed light source. A human eye might gloss over a slightly “off” shadow, but AI flags it as a red alert.
The unnatural blink: humans blink spontaneously, at varying intervals. Deepfake AI, especially older models, often struggled with this, resulting in subjects who blinked too little, too much, or in a strangely synchronised way. Even advanced deepfakes have subtle imperfections in eye reflections or pupil consistency.
The micro-movement: we express emotions through incredibly subtle muscle movements in our faces. While deepfakes can mimic major expressions, they often miss the fleeting, nuanced micro-expressions and the natural fluidity of human movement. AI analyses these minute muscular contractions and the way they affect the skin, spotting where the “performance” isn’t quite authentic.
The tell-tale texture: human skin has pores, imperfections, and a specific texture. AI can sometimes create skin that’s too perfect, appearing unnaturally smooth or lacking the subtle variations of real human skin texture. Hair edges might show slight blurriness or a “pasted-on” quality.
Audio’s unseen footprint: for deepfake audio, AI listens for background noise consistency, the subtle intonation patterns that make speech sound natural, and even the way a voice interacts with a room’s acoustics. Fakes can often have a “flat” sound or lack the natural imperfections of real speech.

No human, no matter how vigilant, can spot these flaws, especially as deepfake technology continues to advance. This is precisely why AI is essential to fight AI.

Tools like VerifyLabs.AI leverage sophisticated algorithms and massive datasets to act as your digital detective, scanning for these invisible tells. We don’t rely on gut feelings; we rely on deep, data-driven analysis to tell you what’s real and what’s a dangerous fabrication.

Equip yourself with the power of AI to see what your eyes can’t.

Gut instinct: trust it before you bust it

July 16th 2025

It’s evening in a corporate office in a major world capital. The hustle and bustle has thinned as colleagues start to go home. An executive sits at their desk, wanting to tie up due diligence before leaving for the nightly commute.

The exec is examining a new client’s details and is uploading a scan of their passport.

It looks fine. The photo is nice and sharp. The layout is clear and all the markings are exactly where they should be.

Nothing about the passport made the exec want to check any further. And the proofs of address and other forms of ID also looked good.

But nevertheless they’re feeling uneasy.

Something the client said on their Zoom call was bothering them.

The client said the weather was sunny, but if they were in London where they alleged they were, they’d have known that it had been pouring with rain for the last two weeks.

In the meeting the exec explained it away thinking they were being ironic, or had made an attempt at humour. But the exec’s tummy feels inexplicably tight and off somehow and, despite being tired, they wonder what to do.

If this were you, would you:

Continue onboarding your client ignoring your bodily dis-ease by rationalising away your feelings as a misunderstanding?

Ask for a robust check on your client’s details, running them through a deepfake detector and asking another human for their opinion?

Our gut-brain connection is a powerful analytics system that often “knows” that further checks are needed before our conscious minds do. When faced with complex decisions where data is incomplete or overwhelming, your gut integrates a vast number of subconscious variables that your logical mind might overlook.

Your gut instinct is not a mystical feeling; it’s a biological and neurological event rooted in four key scientific principles:

The gut-brain axis: your gut contains more than 100 million neurons, forming a “second brain” known as the Enteric Nervous System. This system is in constant, two-way communication with your primary brain via the vagus nerve. A gut feeling is your brain interpreting the massive flow of data—including hormones and nerve signals—coming directly from your gut.
High-speed pattern recognition: a gut feeling is the physical result of your brain’s subconscious processing. It rapidly scans your lifetime of stored experiences and memories for patterns. When it detects a match or mismatch with a past situation, it triggers a physical, visceral sensation long before your conscious mind has had time to logically analyse the situation. It’s a biological “red flag” or “green light.”
A primal survival circuit: this system evolved to ensure human survival by providing immediate risk assessment. The unease or comfort you feel in a situation is this ancient circuit making a snap judgment—”safe” or “threat”—based on subtle environmental cues, helping you react quickly to potential dangers.
Microbiome and neurotransmitters: the trillions of microbes in your gut directly influence your intuition. They produce and help regulate critical neurotransmitters responsible for mood and cognition, including over 90% of your body’s serotonin. The health of your gut microbiome can therefore directly impact the clarity and accuracy of the signals sent to your brain.

Listening to your gut is listening to a powerful form of protective intelligence: a combination of real-time data from your “second brain” and high-speed analysis from your subconscious mind.

There are many accounts of deepfake attacks where victims override their initial bodily intuition, explaining it away.

Listen to your gut if:

You feel something is too good to be true
You feel an initial sense of disbelief or weirdness
You feel that something just doesn’t add up
You see contradictory or non-aligned facts
You spot just one thing that’s off – like a date, a spelling mistake or an usual idiom
You feel someone’s tone of voice doesn’t match their skills and experience
You feel someone’s mode of expression doesn’t fit their purported cultural or ethnic background
You can’t put a finger on what you feel; it just doesn’t feel right somehow.

Always Verify it first

Set an AI to catch an AI

An AI image of a young woman with green eyes generated by Gemini Pro

July 16th 2025

We asked Gemini 2.5 Flash to use everything it knows (including the latest research and common limitations of current generative AI), to tell us how to spot deepfakes that are too good for the human eye to detect.

Gemini had a 3-second think about things that then said that the giveaways often lie in subtle, systemic inconsistencies in physiological and environmental details that betray a lack of genuine understanding of physics and human biology.

Here’s its findings:

Physiological Anomalies (Eyes & Mouth are key):
- Eyes (Reflections, Pupils, Blinking): even when a face looks incredibly real, AI often struggles with the nuanced, consistent reflections in both eyeballs, the perfect circularity/ellipticity of pupils, and realistic, spontaneous blinking patterns (e.g., blinking too little, too much, or unnaturally synchronously). The human visual system is highly attuned to these subtle facial cues.
- Mouth/Teeth: while lip-syncing has improved, the full range of natural mouth movements during speech, especially the appearance of the tongue and the specific shapes of teeth, can still be inconsistent or slightly “off” in high-fidelity deepfakes.
Environmental & Physical Inconsistencies (Beyond the Face):
- Lighting and Shadows: this is a major challenge for AI. Deepfakes often struggle to maintain perfectly consistent lighting and shadow interactions across the entire scene, especially when the subject moves or the light source is complex. You might see subtle discrepancies in how light falls on the face versus the rest of the body or background.
- Hair & Background Artifacts: the edges of hair can sometimes appear unnaturally blurry, “shimmery,” or have a slightly “pasted on” look. Backgrounds may contain subtle distortions, flickers, or lack consistent detail and movement, particularly around the edges of the manipulated subject.
- Micro-expressions and Body Language: AI models are getting better at generating major expressions, but they still struggle to perfectly replicate the incredibly subtle, fleeting micro-expressions and natural, fluid body language that humans unconsciously display. A deepfake might look good in a static frame but feel “off” in motion because these minute details are missing or unnatural.

The reason these are often the “last bastions” of detection for advanced deepfakes is that generating them requires not just replicating pixels, but accurately simulating complex real-world physics, biological processes, and nuanced human behaviour – something current generative AI still finds challenging. Dedicated AI detection tools are trained to spot these specific, often microscopic, anomalies that are invisible to the naked eye.

10 questions to ask before you believe

July 16th 2025

(1) What is the source of the content? Is it from a reputable, known source, or an unfamiliar website or social media account?

(2) Does the context in which the image is presented seem plausible or sensationalist? Are there any accompanying claims?

(3) Are there visible inconsistencies in lighting, shadows, or reflections within the image, particularly around the subject’s face or body compared to the background?

(4) Do the edges of the person or object in question look unnaturally sharp, blurry, or pixelated compared to the rest of the image?

(5) Are there any unusual distortions or artifacts in facial features, such as eyes, teeth, ears, or hair? Do they look symmetrical or natural?

(6) Does the skin texture look overly smooth, waxy, or patchy? Are there any inconsistencies in skin tone or blemishes?

(7) If it’s a known person, does their expression, pose, or the situation depicted align with their known behaviour or public persona?

(8) Are there any oddities in the background details? Do objects appear distorted, or are there any illogical elements present?

(9) Ask: have I seen this image elsewhere? Or can I find other sources corroborating or debunking it using a reverse image search (e.g., Google Images, TinEye)?

(10) Are there any subtle digital artifacts, such as unusual patterns in textures (e.g., hair, fabric), inconsistencies in focus or resolution between different parts of the image, or tell-tale signs of digital “stitching” or “inpainting” that suggest a generative AI process was used?

Newer >