Apate’s AI Agents Go Undercover to Trap Scammers Before Victims Are Hit

Australian startup Apate.ai is using AI agents in an unusual way: to pose as scam victims, keep fraudsters busy, and collect intelligence that banks, telecom operators and government agencies can use to stop attacks earlier. Founded in 2024 from research at Macquarie University in Sydney, the company says its undercover agents have already burned through 17 years of scammers’ time across 2.5 million conversations.

The approach flips the usual AI productivity story on its head. Instead of speeding up sales, support or coding, Apate’s agents answer scam calls, engage with suspicious texts and emails, click through phishing lures, and enter online groups where fraudsters recruit victims. By staying in character, the agents aim to coax out the details that typically surface only after someone has been defrauded.

That intelligence model has attracted major customers, fresh venture capital and attention at TechCrunch Disrupt, where Apate is one of the finalists in the Startup Battlefield 200. As scam operations become more automated and more convincing, the startup is betting that deception can be turned into a defensive tool.

What Apate is actually doing with AI agents

Apate’s core idea is straightforward: let machines waste a fraudster’s time so that defenders can learn faster. The company deploys thousands of agents that behave like plausible victims, each with a distinct persona, and keeps them in circulation across phone, text, email and social platforms.

Those digital decoys can respond to scam calls, continue dubious email threads, follow phishing links, join suspicious groups, and even ask for details in a way that encourages scammers to reveal how a campaign works. The goal is not merely to block a number or report a domain after the fact. It is to gather useful evidence while the fraud is still active.

How the agents stay convincing

Apate says the agents are designed to sound and behave like different kinds of people, with variation in age, gender, accent and language. That flexibility helps keep conversations alive long enough to reveal the scammer’s tactics, preferred payment routes, impersonated brands and infrastructure.

The company says customers can also give the agents specific objectives. For example, an agent might be tasked with coaxing out a bank account number, a crypto wallet address, or a malicious website. In practice, that turns scam baiting into a controlled intelligence operation rather than a hobbyist exercise.

Apate argues that most fraud tools work too late, after a criminal has already reached a customer. The company says its agents go straight to the scammer, collecting information before harm is done and feeding that insight back into anti-fraud systems.

Why the startup thinks its method matters now

The timing is important. AI has made it cheaper and easier for scammers to run larger, more believable operations. Fraudsters can now automate outreach, translate messages instantly, and use synthetic audio or video to impersonate trusted figures.

According to FBI figures cited by Apate, law enforcement received more than 22,000 complaints involving AI-related crime in 2025, with total losses reaching about $893 million. Investment scams alone accounted for more than $632 million of that total, underscoring how quickly AI-enabled deception has become a serious threat.

That shift has made traditional defenses feel reactive. Telecom companies can block suspicious numbers, but fraudsters can rotate to new ones. Banks can monitor transactions, but the scam may already be in motion by then. Apate’s pitch is that earlier visibility into scam tactics can help businesses intervene sooner.

From old-school scam baiting to machine-scale defense

People have long enjoyed wasting scammers’ time, whether by stringing them along in phone calls or turning the interaction into a prank. Apate’s version is more systematic and more commercially focused. Instead of one person spending an afternoon annoying a fraudster, software can run thousands of parallel conversations and turn the results into structured intelligence.

That shift also reflects a broader change in the threat landscape. If scammers increasingly rely on automation, defenders may need automation of their own to keep pace. Apate’s approach is one example of that emerging arms race.

How a single scam call inspired the company

The startup’s origin story began with a family moment rather than a boardroom strategy session. Apate founder and chief executive Dali Kaafar reportedly received a scam call while at a family picnic several years ago and decided to keep the caller on the line for the sake of his children, who found the encounter amusing.

Kaafar managed to hold the call for 44 minutes. What seemed like a joke at the time became the seed of a business idea. If one human could waste that much of a scammer’s day while learning something in the process, Kaafar wondered whether software could do it at scale.

That question later informed research at Macquarie University, where Kaafar and colleagues used machine learning and natural language processing to study scam calls and improve detection methods. The startup grew out of that academic work and the idea that meaningful anti-fraud intelligence could be collected from live interactions instead of post-incident reports.

Who uses Apate’s system today?

Apate says its customer base already includes banks, telecom firms, government agencies and utilities. Among the named users are Commonwealth Bank of Australia and TPG Telecom, two large Australian companies that face high volumes of fraud-related activity.

The company says the system has produced measurable outcomes for some clients. TPG has redirected nearly 590,000 calls to Apate’s agents, according to the startup. Commonwealth Bank has seen a 76% drop in publicly reported scam losses, while another Australian bank using the platform has reportedly disrupted 22 scam campaigns through outbound agent activity.

Those results, if sustained, suggest the company is offering something more than novelty. For regulated industries that lose money not only to direct fraud but also to customer confidence erosion, the promise of better early-warning intelligence is commercially significant.

What information do the agents extract?

Apate says the intelligence it gathers goes beyond simple nuisance value. The agents aim to uncover the practical details that make scam operations work, including bank accounts, crypto wallets, phone numbers, web domains and the names of brands being impersonated.

That information can be used to update blocklists, train detection systems, identify fraud trends and understand how scams are evolving. For organizations trying to protect employees and customers, the difference between a generic warning and campaign-specific intelligence can be substantial.

How the insights portal fits in

The company also offers customers an analytics portal that summarizes scam activity against specific brands, products or user groups. Instead of receiving raw conversation logs only, clients get structured data that can inform operational decisions and anti-fraud campaigns.

This layer matters because many organizations struggle with intelligence overload. A useful anti-fraud platform has to convert messy conversations into something a security team, risk department or operations group can act on quickly.

Why investors are backing the idea

Apate raised $8.15 million in seed funding in August from Lobby Capital, OIF Ventures, Investible, Concept Ventures and Baobab Ventures. The company says it has raised just under $10 million in total and plans to use the money for product development and international expansion.

The startup’s appeal to investors is easy to understand. Fraud is a huge and growing market pain point, and the companies exposed to it often have large budgets and urgent incentives to spend on protection. A tool that helps reduce losses, improve response times and strengthen brand safety can have a clear value proposition.

There is also a broader platform story here. As AI agents become more capable, companies are exploring not only how to use them for efficiency but how to deploy them in adversarial environments. Apate sits in that latter category, where commercial opportunity comes from defending against increasingly automated crime.

How does Apate’s technology work?

Apate says its system uses both frontier AI models and proprietary fine-tuned models. The company trains those models on conversations its agents have had with scammers, along with research derived from Kaafar’s academic work.

That mix is important because scam interactions are highly variable. A useful model needs to understand conversational cues, cultural context, social engineering patterns and the subtle ways a fraudster attempts to build trust or urgency. General-purpose models can provide breadth, while specialized tuning can improve performance in the narrow domain of fraud.

The startup’s text agents are also able to search forums, investment groups and social platforms to identify emerging scam activity. Once they locate a suspicious thread, they can enter the conversation under a fabricated identity and continue probing for signs of malicious intent.

What makes the system different from ordinary fraud detection?

Ordinary fraud detection usually focuses on signals inside a company’s own systems: a suspicious transaction, an unusual account change, or a call flagged by a telecom network. Apate is trying to move the defender closer to the attacker by engaging earlier in the fraud lifecycle.

That difference is strategic. Early intelligence can shape how institutions write warnings, configure controls, prioritize enforcement and educate customers. In a fast-moving scam campaign, a few hours of lead time may matter more than a retrospective report.

Timeline of Apate’s rise

The company’s development path combines academic research, a practical scam encounter and commercial rollout. The sequence helps explain why the startup is drawing attention beyond the typical fraud-tech niche.

What the company says its scale looks like

Apate says its agents have been active across 2.5 million conversations and have collectively consumed the equivalent of 17 years of scammer time. Those are hard numbers to independently verify from the outside, but they reflect the scale at which the company says it is operating.

Scale matters in fraud prevention because scam campaigns often depend on volume. The more contacts a fraudster can place, the higher the odds of finding a vulnerable target. A counter-system that can hold thousands of conversations simultaneously could meaningfully disrupt that economics.

Joffe, the company’s co-founder and chief commercial officer, said Apate’s agents speak to tens of thousands of scammers every day across calls, messages, email and the web. He described the company’s output as intelligence derived from direct interaction with fraudsters, rather than speculation or delayed reporting.

Joffe said the business is built around moving upstream: instead of trying to stop a scam only after a customer has been exposed, Apate tries to collect information from the scammer first and hand that insight to customers in usable form.

What comes next for Apate?

The startup is now preparing for growth beyond Australia. It plans to open a London office and increase its U.S. footprint, signaling that it sees demand for anti-fraud intelligence outside its home market. That expansion will likely test whether its approach translates across regulatory regimes, scam styles and telecom environments.

International growth could also bring competition. The fraud-tech sector is crowded with tools for identity verification, transaction monitoring, domain takedowns and call screening. Apate’s challenge will be to prove that its AI-driven undercover model creates a defensible edge rather than just a clever feature.

Still, the company has one advantage many startups would envy: a product story that is immediately understandable. In a world where scammers increasingly use AI, Apate uses AI to waste their time, learn their tactics and help institutions respond before the damage is done.

Why this story matters

Apate’s emergence is part of a broader shift in how AI is being used. The public conversation often focuses on generative tools that write code, create images or answer questions. But the same underlying technology can also be used for defense, deception detection and adversarial intelligence.

That dual use is especially relevant in fraud, where the cost of missing a signal can be measured in stolen savings, compromised accounts and shattered trust. If Apate can keep proving that undercover AI agents deliver actionable intelligence, it could help define a new category in the fight against financial crime.

For now, the company’s most unusual metric may also be its most telling: every frustrated curse from a scammer is treated as a small victory.