The 'Strict Father' of Northern Myanmar: This Company Deploys 200,000 "AI Best Actors" to Strike Back Against Scammers

Let AI go undercover, and drive the scammers crazy.

Written by: Han Paopao | TechFlow

Inside an office in Sydney, every time foul language and angry roars spill through the speakers from telecom fraudsters, the entire team erupts into cheers.

“We treat this as a quantifiable metric of scammer frustration,” laughed Brad Joffe, co-founder and Chief Business Officer at Australian cybersecurity startup Apate.ai. “Whenever a scammer starts hurling abuse over the phone, we celebrate. We’re probably one of the very few companies in the world that gets happy when users curse at our product.”

While most tech companies are busy using AI Agents to help humans write code, book flights, and generate reports to boost productivity, Apate.ai—selected for the Startup Battlefield 200 at the 2026 TechCrunch Disrupt pitching contest—has given AI a mission that sounds delightfully mischievous yet incredibly satisfying: go undercover, actively hunt scammers, and drive them crazy.

To date, this around-the-clock army of “AI victims” has conducted over 2.5 million conversations with fraudsters worldwide, cumulatively draining scammers of a full 17 years of operational time.

This isn’t merely a prank for venting frustration.

Behind those segments of grating nonsense that drive scammers up the wall, a multi-million-dollar anti-fraud business is quietly humming along.

A 44-Minute Picnic Call

The story begins with a family picnic brutally interrupted.

In November 2021, Dali Kaafar, a cybersecurity professor at Macquarie University in Sydney, was enjoying some outdoor time with his family when his phone suddenly rang with an unknown number. The caller claimed to represent a well-known tech giant and used a polished script in an attempt to scam him.

An ordinary person would likely have hung up within a second. But Kaafar, who studies cybersecurity, got a bright idea and decided to entertain the kids nearby. He posed as the "perfect prey"—clueless about electronics, slow to react, and overly polite—and chatted meanderingly and awkwardly with the scammer on the phone, dragging the call out for a solid 44 minutes.

The kids around him burst into laughter, and the scammer on the other end finally snapped and hung up. But once the giggles subsided, Kaafar keenly realized two things:

First, during those 44 minutes, this scammer was completely tied up on the phone line, unable to target any actual potential victims;

Second, he had sacrificed 44 precious minutes of his own family time.

“Teasing a scammer occasionally counts as entertainment, but if humans actually did this every day, it becomes hard to say who’s wasting whose time.”

An idea took shape in his mind: What if thousands of tireless AI bots were trained to take these calls on behalf of humans, draining scammers of their time while casually extracting their criminal blueprints?

That concept quickly evolved into a major research initiative at Macquarie University. Led by Kaafar, the team collected massive volumes of real scam calls, phishing emails, and social media chat logs. Using machine learning and natural language processing, they deconstructed the scammers’ "social engineering" tactics to build a dialogue system capable of maintaining a realistic persona over extended periods.

In 2024, the project officially spun off from the university, adopting the name Apate. In Greek mythology, she is the goddess of deception and trickery who emerged from Pandora’s box.

An anti-fraud company putting the "goddess of deception" right on its signboard.

200,000 “AI Best Actors”

To keep wary, professional scammers on the phone line for dozens of minutes, the AI can’t just mechanically answer questions. It must know how to exploit human greed and impatience.

At its core, telecom fraud is a funnel screening process: for a scammer to get a stranger to click a suspicious link and bypass banking security controls to transfer funds, there inevitably must be a long process of explanation, coaxing, and verification.

As long as the prey shows signs of “I have money, I’m scared, I want to cooperate, but I’m just clumsy and don’t know how to operate it,” the scammer will absolutely refuse to hang up.

Apate’s AI agents take this performance of “push-and-pull hope” to the extreme.

The company has built thousands of virtual personas with varying ages, genders, accents, languages, and even cultural backgrounds. For example, an elderly bot named Malcolm, speaking with an authentic British accent, might lethargically ask after hours of the scammer laying out a mysterious financial crisis: “Sorry mate, what exactly was that service you mentioned?”

They pepper their speech with Australian slang and dry humor, pretending over the phone to have bad reception, to lose their reading glasses, to mistype keys on the keyboard, or to show perfectly timed panic at popup alerts.

More impressively, this squad of AI undercover agents doesn’t just sit back and wait.

Telecom providers seamlessly route identified high-risk fraudulent calls directly into Apate’s systems. Meanwhile, Apate has planted hundreds of thousands of “honeypot” numbers across communication networks, specifically designed to bait bulk SMS and auto-dialer systems.

Furthermore, they also proactively strike.

Text-based AI agents patrol forums, Telegram investment groups, and social media platforms 24/7 like real netizens. Once they detect signs of a pig-butchering scam or fake investment schemes, they proactively click phishing links, add friends, join groups, and engage in deep private chats with scammers, masquerading as investors desperate for quick wealth.

During a pilot program with Australian telecom giant TPG Telecom, a single AI-to-scammer call hit a staggering record of 54 minutes. The scammer thought they were one step away from accessing the victim’s pension, unaware that even the breathing sounds on the other end were synthesized by computing power.

Scammers Provide Intel, Banks Pay for Data

If Apate merely kept scammers busy doing nothing, it would barely qualify as a public-interest gag and couldn’t sustain a commercial company. What truly gets venture capital firms and banking giants to part with real money is the “golden intelligence” effortlessly harvested amidst the rambling nonsense.

“The entire anti-fraud industry has almost always been acting too late,” pointed out co-founder Brad Joffe. “Telcos block numbers, but scammers can swap to a new one for a few cents. Banks monitor abnormal transactions, but by the time alarms sound, scammers have already brainwashed the clients and often transferred the money.”

Under Apate’s business model, however, the defensive line is pushed straight to the doorstep of the scammers’ lair.

No matter how elaborate the scammer’s script, if they want to get paid, they’ll inevitably have to show their hand at the final stage: providing a bank account number to receive funds, a cryptocurrency wallet address for laundering, a download link for remote control software, or a spoofed phishing website domain.

During these conversations, Apate’s agents have only one core KPI: to coerce scammers into spilling these critical infrastructure details.

The moment a scammer drops a bank account or USDT wallet address in the chat, Apate’s backend system extracts and structures it within seconds, then pushes it directly via its proprietary real-time intelligence platform (Apate Insights) to banks, telcos, and government regulators.

Clients can even issue “mission orders” to the AI undercover agents through the backend. For instance, targeting a recent surge in impersonation scams involving well-known brands, they can deploy specific AI personas to infiltrate and conduct targeted investigations, digging deep into the underlying fund pool addresses.

This effectively means that before scammers can even lay a finger on real humans, the money-mule accounts and phishing websites they’re using to collect cash are already blacklisted across major banks’ risk-control systems.

Apate’s operational data is highly impressive:

• Commonwealth Bank of Australia (CommBank): After integrating Apate’s real-time fraud intelligence into its risk-control system and combining it with multiple internal anti-fraud defenses, the bank reported a 76% drop in publicly disclosed customer fraud losses compared to peak levels.
• TPG Telecom: Has cumulatively routed nearly 590,000 fraudulent calls directly into Apate’s AI undercover network, achieving a 98% accuracy rate in detecting scam traffic.
• Proactive Cleanup: Another unnamed major Australian bank leveraged Apate’s outbound AI agents to launch counter-offensives, directly crippling 22 active large-scale scam syndicate operations.

The capital markets quickly sniffed out the value.

In August 2026, this lean 20-person team announced the closing of an $8.15 million (approx. 11.4 million AUD) seed funding round. Backers included prominent firms such as Lobby Capital, OIF Ventures, and Investible, bringing their cumulative financing close to $10 million.

Today, they are planning to open an office in London and are making a major push into the U.S. market.

When Scammers’ AI Meets Anti-Fraud AI

The rise of Apate reflects a harsh reality on today’s global cybersecurity battlefield: the telecom fraud industry has long since completed its own “AI arms race upgrade.”

According to the latest data released by the Federal Bureau of Investigation (FBI), the agency alone received over 22,000 AI-related crime complaints in 2025, resulting in direct economic losses of $893 million. Over $632 million of that came from AI chat-driven investment scams (pig-butchering schemes) and scams using Deepfakes to clone celebrities and executives’ voices and faces.

In the past, a single worker in a fraud compound would smoke from their throat after dialing dozens of calls a day. Now, criminal syndicates are using LLM-powered automated scripts to conduct mass parallel outbound calls and auto-replies in the tens of thousands. Relying on traditional human volunteers to “tease scammers” is no longer enough; only computing power can fight computing power.

In a Chinese context, Apate earns itself a nickname: the stern father of Northern Myanmar.

Since scammers are using AI to mass-hunt for victims, while anti-fraud companies are using AI to mass-disguise as victims, could we soon witness a massive “silicon-based proxy war” playing out across global communication cables and social networks?

Thinking further down that line, the picture grows somewhat absurd.

In a late-night encrypted chat room, the fraud compound’s AI is passionately laying the groundwork for fake romance and high-yield crypto mining pools, while the opposing Apate AI is eloquently portraying a wealthy but bewildered retired grandfather.

The two emotionless large models volley back and forth, chatting for three days and nights, burning through GPUs and electricity costs in each other’s server rooms, while humans peacefully sleep beside them.

As for this seemingly absurd endpoint, founder Dali Kaafar actually contemplated it years ago. His response was crisp and resolute:

“If in the end, all the scam bots in the world are busy chatting with our anti-fraud bots, allowing real humans to avoid the ruin of losing everything, then I’d gladly consider it a complete victory.”

Join the official Coincamps community:

X: https://x.com/coincamps

Telegram: https://t.me/coin_camps