πŸ“° Key Summary

Apate.ai is an Australian startup founded in 2024, spun out of research at Macquarie University in Sydney. Its core approach is deploying large numbers of AI agents that pose as scam victims and proactively engage with fraud rings. These AI agents answer scam calls, reply to suspicious messages and emails, click phishing links, and even join social media groups to lure scammers in. They can mimic people of different ages, genders, and accents, and converse in multiple languages β€” all with the goal of dragging out the interaction with scammers as long as possible in order to extract intelligence.

The company says its AI agents have already burned through 2.5 million conversations, costing scam operations a combined 17 years of time. For Apate’s clients β€” banks, telecoms, government agencies, and utility companies β€” what matters even more is that these undercover AI agents work to trick scammers into revealing information that would normally only surface after a real victim has actually been defrauded: bank accounts, crypto wallet addresses, phone numbers, malicious websites, impersonated brand names, and other details of the scam playbook. Apate then hands this intelligence back to its clients to strengthen their anti-fraud defenses.

The article notes that AI has massively amplified the scale of scams and hacking: the FBI received more than 22,000 AI-related crime complaints in 2025, resulting in roughly $893 million in losses, with over $632 million of that coming from investment scams β€” scammers using AI to hold conversations and deepfake the voices and appearances of celebrities, executives, and other trusted figures.

Apate founder and CEO Dali Kaafar got the idea from a real scam call he received β€” he deliberately played dumb and strung the scammer along for 44 minutes, which sparked the idea of automating that behavior and harvesting intelligence from it, eventually growing into a research project at Macquarie University. The company is a finalist in this year’s TechCrunch Disrupt Startup Battlefield 200 competition.

πŸ’¬ JudyAI Lab’s Take

Apate.ai has AI agents pose as scam victims, answering scam calls, replying to suspicious messages, and joining social media groups to lure scammers in β€” racking up 2.5 million conversations that have cost scam operations a combined 17 years. This “use AI to fight AI” defense strategy is worth paying attention to.

This case reflects a shift: anti-fraud efforts used to be mostly about detecting and warning after the fact. Apate turns AI agents into active bait instead β€” simulating different ages, genders, accents, and even multiple languages, not to immediately expose the scam, but to drag out the conversation as long as possible and extract structural intelligence from it: bank accounts, crypto wallet addresses, and details of the scam methodology. FBI data shows over 22,000 AI-related crime complaints in 2025, with losses nearing $893 million β€” investment scams accounted for the largest share, and scammers themselves are using AI deepfakes of voices and appearances to win victims’ trust. When both attackers and defenders treat AI as a tool for scaling up, the design question shifts from “is this real or fake, right now” to “how do we trade conversation length for intelligence.”

For AI builders, this points to a direction worth extending: when evaluating an AI agent, don’t just look at whether its answers are accurate β€” also consider how long it can sustain an interaction, and what it can gather along the way.

πŸ“… Original Source

πŸ”— Further Reading