An Australian startup, Apate, has built an army of nearly 200,000 AI personas that pose as gullible victims to bog down and gather intelligence on online scammers. The bots can conduct convincing phone calls and messaging conversations across platforms like WhatsApp and Telegram, deliberately stringing fraudsters along so each hour wasted is an hour the criminals aren’t targeting real people.
Apate’s founder, Dali Kaafar, says the company even tracks how often frustrated scammers curse at their faux victims as a performance metric. The operation now handles “hundreds of thousands” of calls and chats daily. For example, in a six-week period ending in late 2025, Apate’s bots made 600,000 scam calls for Australian telco TPG — which the company says translated into more than 500 days of scammers’ time wasted and an estimated US$13 million in avoided losses.
Origins and scale
The idea began in 2021 after Kaafar entertained his children by stringing a scammer along for 44 minutes during a family picnic. While a professor at Macquarie University, he and doctoral students researched automating that approach. With funding from the Office of National Intelligence, the pilot work moved out of the university and into Apate, spun off in 2023. Today the firm works with major banks and telcos across Australia, the UK, South Africa and Southeast Asia.
Apate started with about 120 personas and has expanded to roughly 197,000 distinct profiles: different ages, genders, accents, vocal tics and realistic hesitations. Models were trained on many hours of recorded interactions between human baiters and scammers so the bots can adopt counter-strategies and keep conversations credible even when a scammer grows skeptical.
Intelligence, not just disruption
Beyond wasting time, the bots aim to collect actionable intelligence for banks, telcos and crypto investigators. In the crypto space Apate supplies wallet addresses and transaction patterns that help tracing firms and exchanges detect where stolen funds are being routed. Kaafar says the bots can extract new wallet addresses by the hundreds or thousands, giving defenders leads before money moves out of reach.
In July Apate reported uncovering a marketplace in India that brokers verified bank accounts to scammers and offers commissions paid in USDT. These kinds of revelations help financial institutions and law enforcement shut down money-collection points used by scam call centers, which Kaafar describes as organised, corporate-style operations.
The wider context and arms race
Apate is not alone in using AI to frustrate scammers. For instance, UK carrier O2 ran an “AI Granny” campaign that tied callers into long conversations about her many cats as a public-awareness stunt. At scale, however, Apate’s roster and data focus set it apart.
At the same time scammers are adopting AI: Apate’s research estimates 20–30% of scam text conversations already use AI. Kaafar argues defenders have a theoretical advantage because their bots seek intelligence, whereas scam bots try to induce actions, making attacker models easier to probe and learn from. He believes that as scammers grow more sophisticated, a defensive AI ecosystem that extracts and shares intelligence will improve the odds of stopping fraud.
Why it matters
Scamming is a vast global industry. Estimates put annual losses in the hundreds of billions to over a trillion dollars worldwide, and organized call centers and broker networks amplify that harm. Apate’s approach combines disruption with evidence-gathering — wasting criminals’ time while feeding banks, exchanges and investigators the leads they need to block accounts, trace funds and dismantle fraud infrastructure.
That mix of scale, forensic value and automation is what Apate says differentiates its service: not just annoying scammers, but turning conversations into preemptive intelligence so victims and institutions can act before funds disappear.
