Apate deploys undercover AI agents to gather scam intelligence

Australian startup Apate says it deploys thousands of AI agents that pose as plausible scam victims in order to gather fraud intelligence. According to TechCrunch, the company was founded in 2024 on research conducted at Macquarie University in Sydney. Its agents can answer scam calls, reply to suspicious messages and emails, follow phishing links, and enter social media or investment groups where scam activity may be taking shape. The stated goal is not only to waste scammers’ time, but also to extract details that organisations can use to prevent fraud earlier in the attack chain.

The company’s pitch is that much anti-fraud work happens too late. In TechCrunch’s account, Apate co-founder and chief commercial officer Brad Joffe contrasted the model with telecom operators blocking numbers after the fact or banks monitoring transactions once contact with a customer has already happened. Apate instead says it goes directly to the scammer. Its agents are presented as fabricated identities with different ages, genders, accents and languages, designed to keep conversations going long enough to surface useful information.

How the system is supposed to work

Across calls, texts, email and the web, Apate says its bots engage fraudsters in real time and then structure what they learn for customers. TechCrunch reports that the information sought can include bank accounts, crypto wallet addresses, phone numbers, malicious websites and the brands being impersonated. On its own website, Apate frames the same process as a way to expose accounts, infrastructure and tactics that scammers depend on. The company says this can help fraud teams disrupt operations before they reach customers.

Apate also says customers can task the agents with specific missions. TechCrunch reports that clients may direct them to obtain a bank account number or a crypto wallet address. The company website expands that logic into several use cases: identifying and tracking mule networks, detecting phishing and impersonation campaigns, supporting real-time fraud operations, and producing evidence for reporting and compliance. In these descriptions, Apate presents its output as structured, timestamped and exportable intelligence that can plug into existing fraud or security workflows.

• Engaging suspected scammers across phone, SMS, email, web and social channels
• Using fabricated identities to prolong conversations and gather operational details
• Turning those exchanges into intelligence on payment routes, scripts, phone numbers and websites
• Feeding that intelligence into customer investigations, takedowns and reporting processes

The company says scale is a key part of the proposition. On its website, Apate argues that manual scam-baiting and human-led intelligence gathering are expensive, inconsistent and hard to expand as fraud volumes rise. Its claim is that AI can take activity that might otherwise happen in dozens of manual engagements and turn it into thousands of simultaneous interactions across languages, channels and time zones.

What Apate says it has seen so far

TechCrunch reports that Apate says its agents have so far consumed 17 years of scammers’ time across 2.5 million conversations. Those figures come from the company’s own claims. The same article says telecom customer TPG has diverted nearly 590,000 calls to Apate’s agents, that Commonwealth Bank of Australia has seen a 76% drop in publicly reported scam losses, and that an unnamed Australian bank has used Apate’s outbound agents to disrupt 22 scam campaigns. The source attributes those statements to Joffe.

Some of those performance claims should be read carefully. TechCrunch reports them as statements from Apate rather than independently verified results, and the available source material does not provide a methodology linking all outcomes directly to Apate’s system. In particular, the claim about Commonwealth Bank of Australia refers to a drop in publicly reported scam losses, but the source text does not detail causation. A cautious reading is that Apate says its customers are using the intelligence to inform anti-fraud measures, and that the company associates those measures with concrete results.

TechCrunch also places the startup in a broader context: scammers themselves are increasingly using AI. The article cites FBI figures saying the agency received more than 22,000 complaints of AI-related crimes in 2025, with about $893 million in losses, including more than $632 million from investment scams. Within that framing, Apate’s model looks like an attempt to use automation not just for detection, but for active intelligence collection inside scam conversations.

Customers, product scope and stated safeguards

On customers, the clearest names in the sources are Commonwealth Bank of Australia and TPG Telecom. TechCrunch also says Apate serves other utilities, telecom companies and government agencies. The company website repeatedly positions the offering for fraud and intelligence teams inside financial institutions and telecommunications operators, with benefits extending to regulatory reporting, partner collaboration and live investigations. It also suggests that diverted scam traffic can become a source of intelligence rather than merely something to block.

As for safeguards, the sources describe them mostly in operational rather than ethical terms. Apate says scam calls can be diverted “safely” into “controlled” or “isolated” environments so that customers are shielded while scammers are engaged. The website also says this can happen without disrupting legitimate calls, network performance or customer experience. In addition, it describes the resulting intelligence as structured and timestamped, creating an audit trail that institutions can use to demonstrate proactive action to regulators.

That said, the available texts do not set out detailed guardrails on matters such as human oversight thresholds, escalation rules, model error handling or limits on impersonation by the bots. The declared guardrails in the sources are therefore relatively narrow: isolation of suspicious traffic, protection of end users from direct exposure, and audit-ready records for compliance and oversight. Anything beyond that would go beyond the material provided.

Research roots and funding

According to TechCrunch, the company’s origin story began with founder and chief executive Dali Kaafar keeping a scam call going for 44 minutes at a family picnic, then asking whether that process could be automated in a useful way. The publication says that idea later became research at Macquarie University using machine learning and natural language processing to uncover scam call content and improve detection and prevention. Apate says it now uses both frontier models and its own fine-tuned models, trained on conversations its agents have had with scammers as well as Kaafar’s earlier research.

Investors have backed that thesis. TechCrunch reports that Apate raised $8.15 million in seed funding in August from Lobby Capital, OIF Ventures, Investible, Concept Ventures and Baobab Ventures. Joffe told the publication that the 20-person company has raised just under $10 million in total. According to the same source, the latest funding is intended for product development and international expansion, including an office in London and a larger presence in the United States.

For English-speaking companies, the practical change described by these sources is a shift from mainly blocking and reacting to scams toward feeding real-time, conversation-level intelligence into fraud operations. Banks, telecom operators, utilities and public-sector bodies could use that information to spot impersonation campaigns earlier, enrich investigations, support takedowns and produce compliance records. Based on the material available, Apate’s promise is not that scams disappear, but that organisations gain earlier visibility into how scam networks operate and more actionable evidence before customers or employees are harmed.