Last updated: August 5, 2026 · By Vishal Swami, Founder & Lead AI Reviewer, AISagely
AI fuels more than half of cybercrime in Africa, Interpol says. Its new African Cyberthreat Assessment Report 2026, released August 3, 2026, found that 55% of reported cybercrime across 36 surveyed countries is now AI-enabled.
Short answer: Interpol's 2026 report found AI linked to 55% of reported cybercrime across 36 African countries. Reported losses more than doubled since 2024, from $192 million to $484 million. AI is doing the grunt work: writing convincing phishing emails, cloning voices, and generating deepfakes for sextortion, at a scale human scammers alone couldn't hit.
I write about AI tools for a living. That means I spend as much time testing the ones built to catch fraud as the ones used to write a blog post. This report matters outside Africa too. Interpol says the same AI-generated Business Email Compromise scams built there are aimed at victims in Europe and North America. The tooling isn't exotic. It's the same voice-cloning and text-generation software you can sign up for this afternoon, just pointed at a target list instead of a to-do list.
What Interpol's report actually found
Interpol's cybercrime director, Neal Jetton, put it plainly: "AI is automating every stage of a cyberattack from reconnaissance and phishing to extortion and evasion," according to Interpol’s own announcement. The report surveyed 36 African member countries. It found cybercrime has shifted from scattered individual scams into what Interpol calls an "industrialised, borderless ecosystem." Losses climbed from $192 million in 2024 to $484 million this year. Scam centers, physical operations where people are trafficked or hired to run scripted fraud, were confirmed in 72% of the countries surveyed.
The AI-specific numbers are the headline. Digital sextortion using AI-generated deepfakes was one of the fastest-growing categories. Interpol's cybersecurity partner TrendAI logged over 600,000 sextortion detections during the review period, Infosecurity Magazine reported. Business Email Compromise fraud is now harder to catch too. A scammer impersonates an executive or vendor to redirect a payment, and the emails read like they were written by the person being impersonated. Interpol attributes that directly to AI-generated correspondence. Ransomware strains including Qilin and Akira targeted hospitals, power utilities, and schools. Criminals are also using AI-generated synthetic identities to get past biometric checks, the same face-and-fingerprint systems banks and government portals use to confirm you're a real person.
It isn't all bad news. Four coordinated law enforcement operations recovered more than $100 million and led to over 1,500 arrests. Seventeen African nations passed or updated cybercrime legislation in the past year. Operation Red Card 2.0 alone clawed back $4.3 million. Interpol's framing is that enforcement is catching up, just not as fast as the tooling is spreading.
What you'll need to protect yourself
You don't need enterprise security software here. You need habits. A phone or laptop with a lock screen, two-factor authentication on email and banking, and ten minutes to review your accounts is enough to start. If you run a small business that handles wire transfers or vendor payments, add one more thing: a callback verification rule for any payment request that arrives only by email. No exceptions, no matter how convincing the message reads.
Step-by-step: spotting an AI-run scam before it works
1. Slow down on urgency
Every category in Interpol's report, sextortion, BEC, romance scams, relies on pressure to act before you think. A message demanding payment, a password reset, or a wire transfer within the hour is a bigger red flag now than a typo ever was. AI has mostly fixed the typos.
2. Verify voice and video out of band
Voice cloning needs only a few seconds of someone's real audio. It can come from a video call, a voicemail greeting, or a public interview. If you get an urgent call from a "family member" or "boss" asking for money or credentials, hang up. Call them back on a number you already have saved, not one given to you in the message.
3. Check for AI-generated correspondence tells
In my testing against real BEC-style samples, AI-written scam emails are grammatically clean but oddly generic on specifics. They get the job titles and company names right, but they're vague on internal details a real colleague would casually reference. If an "urgent" email from your CEO doesn't mention anything only your CEO would know, that's the gap to lean on.
4. Don't trust a face on a video call by default
Interpol's report ties deepfakes directly to sextortion and impersonation scams. Tools like Qualcomm's Halo by Scam AI now run on-device deepfake detection for live video calls, flagging synthetic faces in real time. That alone tells you how mainstream this threat has become.
5. Treat biometric checks as one layer, not proof
Synthetic identity fraud is explicitly called out in the report as a way criminals bypass biometric verification. If a service only asks for a face scan or a fingerprint to move money, that's a weaker gate than one that also confirms your identity through a second channel you control.
6. Report it, even if nothing was lost
Interpol's recovered $100 million and 1,500 arrests came from reported cases. If you spot an AI-generated scam attempt, report it to your bank, your platform, or your country's cybercrime unit. Under-reporting is part of why the loss numbers keep climbing.
Example prompts you can copy
If you want to stress-test whether you'd catch one of these scams, run these through an AI assistant yourself. Compare its output against what actually landed in your inbox.
- Spot the tell: "Here's an email I received asking for an urgent payment [paste text]. List anything that reads like generic AI-written business language rather than something a real colleague would write."
- Practice verification: "Write a two-sentence callback script I can use to verify a wire transfer request by phone before I approve it."
- Check a link or request: "I got this text message asking me to confirm my bank login [paste text]. What urgency or pressure tactics does it use, and what would a legitimate bank message look like instead?"
Common mistakes to avoid
The mistake I see most often: assuming a scam is obvious because a machine wrote it. The opposite is true now. AI-generated messages are cleaner than the human-written scams they replaced. Second is treating video calls as unfakeable proof of identity. Deepfake tools are cheap and fast enough now that a familiar face on a screen isn't verification on its own. Third is skipping the callback step because a request "sounds urgent enough to be real." That urgency is exactly what the scam is designed to create. Fourth, for small businesses specifically: having a payment-verification policy that exists on paper but isn't followed once someone's in a hurry.
Interpol's 2026 African Cyberthreat Assessment: the numbers
| Metric | 2024 | 2026 report |
|---|---|---|
| Cybercrime linked to AI | Not separately tracked | 55% of reported cases |
| Reported losses (annual) | $192 million | $484 million |
| Countries with confirmed scam centers | Not tracked at this scope | 72% of 36 surveyed |
| Sextortion detections (TrendAI) | Not tracked | 600,000+ |
| Recovered funds (4 operations) | — | $100 million+ |
| Arrests | — | 1,500+ |
Tools that make this easier
You don't have to become a security researcher to stay ahead of AI-run scams. It helps to know which tools are built for detection versus generation. My Halo by Scam AI review covers the on-device deepfake detector Qualcomm shipped for live video calls, the closest thing to a consumer answer to the sextortion numbers in this report. If you want to understand the writing tells scammers' AI tools still leave behind, my guide to how to spot AI writing walks through the same checks I use on suspicious emails. Small businesses handling vendor payments should read my best AI tool for small business guide before adding any AI tool to a finance workflow. The wrong integration can widen your attack surface instead of shrinking it. For a sense of how convincing a cloned voice can sound, my how to use ElevenLabs guide covers the same voice-generation category criminals are misusing. And if you're building a fraud-checking habit, my free AI tools roundup and AI tool ratings guide are good starting points for vetting a tool before you trust it with anything sensitive.
My take
The number that stuck with me isn't 55%. It's that reported losses more than doubled in a single year while arrests and recovered funds also went up. Both things are true: enforcement is working, and criminals are still winning the growth race. The AI tools that write the phishing email or clone the voice cost less and scale faster than the humans chasing them down. Interpol's report is about Africa specifically, but the tooling doesn't respect borders. Neither do the BEC scams it describes hitting targets in Europe and North America. Treat every unexpected urgent request, by email, call, or video, as unverified until you've checked it through a channel the scammer doesn't control.
Frequently Asked Questions
What percentage of cybercrime in Africa is linked to AI?
Interpol's African Cyberthreat Assessment Report 2026 found that 55% of reported cybercrime across the 36 African countries it surveyed is now AI-enabled, covering everything from phishing to deepfake-driven sextortion.
How much money has been lost to AI-driven scams in Africa?
Reported cybercrime losses rose from $192 million in 2024 to $484 million in the 2026 report, a jump Interpol attributes largely to AI-assisted scams, credential theft, and automated social engineering.
Is this only a problem in Africa?
No. Interpol's report notes that Africa-based Business Email Compromise operations target victims in Europe and North America. The AI tools involved, text generation and voice cloning, are available worldwide, not region-locked.
How can I tell if a call or video is a deepfake?
Hang up and call the person back on a number you already have saved, not one given to you in the message. Tools like Qualcomm's Halo by Scam AI now flag synthetic faces on live video calls in real time. Grammatically clean but detail-vague messages are a common tell for AI-generated text.
Did enforcement improve at all?
Yes. Four coordinated operations recovered more than $100 million and led to over 1,500 arrests. Seventeen African countries updated their cybercrime laws in the past year. Losses still grew faster than recoveries, according to Interpol.