As AI fuels more sophisticated scams, it is also being deployed on the defensive side. One notable effort comes from Apate, an Australian company whose AI bots pose as potential victims to tie up scammers' time. Founder Dali Kaafar says a minute a scammer spends talking to a bot is a minute they are not reaching real people, especially since many scammers use automated dialing tools.
Apate's platform, used by banks and supported by telecom companies, operates roughly 350,000 bots that not only answer calls but also infiltrate scam chat groups and reply to texts. The bots have varied personalities, languages, and profiles to seem convincing, and they collect intelligence in the background. The company reports more than 250,000 pieces of real-time data, from fraudulent URLs to bank details and money mule accounts.
Separately, researchers are improving honeypots with large language models. Mark Vero of ETH Zurich and colleagues found that LLM-powered honeypots kept AI agents attacking their system significantly longer than predictable ones, and the attackers were less likely to identify the decoys. This suggests generative AI can help defenders waste attackers' resources and gather insights.
The article notes that cybercriminals still launch billions of messages and calls each year, and no single tool will stop them. But AI-driven countermeasures are emerging as a practical business for security vendors, offering banks and telecoms a way to disrupt fraud while collecting actionable intelligence. The source does not present competing views, as it is a single report on these developments.