Australian company Apate develops and operates AI bots that prolong phone calls and message exchanges with scammers. The bots pose as victims, consuming the scammers’ time while collecting information such as scam website URLs and payment account details.
Keeping Scammers on the Phone to Waste Their Time
The safe response for people is to hang up immediately when a call seems suspicious. Apate’s AI takes the opposite approach. It does not fall for the scam, but shows enough interest to keep the other party trying to make a pitch, prolonging the conversation.
The aim is to use up time scammers would otherwise spend on their next target. Unlike measures that block incoming calls to protect users, this approach seeks to disrupt the work of those carrying out the scam. However, the time spent prolonging a conversation does not, by itself, show how much harm has been prevented.
Founder and CEO Dali Kaafar says calls with the bots can last more than two hours. The bots have different personalities, language abilities, and backstories. They do not always answer the phone; sometimes they say they will talk later and hang up. Varying their responses is one way to make them harder to identify as AI.
A Conversation Designed to Be Skeptical but Keep Listening
The key is not to play someone who believes everything. Instead, the bots express suspicion without ending the conversation, leaving scammers room to think they can still persuade them. The design keeps the other party explaining by moving between wariness and interest.
When WIRED tried a public demo while playing the role of a scammer, it spent about six minutes pitching a fictional cryptocurrency investment, but the AI did not agree to invest. The test found that the timing of the responses and the flow of conversation felt natural. However, this was a demo experience, not a trial measuring how much real-world fraud the system can prevent.
Collecting Scam Websites and Payment Details Through Conversation
Another goal is gathering information. If the other party continues the pitch, details used to carry out the scam—such as the URL they direct victims to or the account for a bank transfer—may come up in the conversation. The exchange turns more than a way to waste time into material for investigations.
Kaafar says the operation involves about 350,000 bots and has collected more than 250,000 pieces of information. This includes scam website URLs, accounts used to move money, and banking details. These are figures provided by the company, not the results of independent verification.
The operation is not limited to phone calls. The bots respond to text messages and are also said to infiltrate online chat groups used by scammers. Banks use the service, and it also receives support from telecommunications companies. However, the materials provided do not identify specific customers or say how many accounts or websites have been shut down using the collected information.
AI That Keeps Attackers Engaged Also Works With Fake Systems
Deliberately luring in attackers and observing their behavior is a long-standing cybersecurity defense technique. A common example is a honeypot: a system made to look real that lets defenders study intruders’ methods or keep them occupied.
Mark Vero, a doctoral researcher at ETH Zurich, and colleagues are studying honeypots that incorporate large language models for tasks such as generating text. In the research they described, AI agents carrying out attacks stayed engaged longer, and were less likely to identify the honeypot as a decoy, compared with systems whose responses were more predictable.
While Apate generates human-like conversations, this research generates responses from systems that appear genuine. What they have in common is using AI not simply to detect suspicious behavior, but to engage directly with the other party. However, the study focused on AI attackers, and its findings cannot be equated with effectiveness against human scammers.
A Track Record of Long Calls Is Not the Same as a Record of Reducing Harm
Call duration and the number of bots are not enough to evaluate this approach. It is necessary to separately assess how much it disrupts scammers’ work, whether the information collected leads to accounts or websites being shut down, and whether harm is ultimately reduced.
If banks, telecommunications companies, social media platforms, and law enforcement agencies can share information, clues gathered in conversations could potentially support other countermeasures. On the other hand, scammers also use large-scale automated calling and organized operations, so drawing some of them into long calls may not be enough to stop them altogether.
The materials provided document Apate’s operating scale, the number of items of information collected, and examples of lengthy conversations. They do not include data showing a reduction in financial losses or how much the scam success rate has fallen.
Comments