An Australian cybersecurity startup is drawing fresh attention in the United States for using artificial intelligence to impersonate potential fraud victims, engage suspected scammers and extract information about criminal operations. Apate.ai says its network of automated agents has conducted more than 2.5 million conversations, collectively consuming approximately 17 years of scammers’ time. The company’s participation in San Francisco’s TechCrunch Disrupt startup competition this week has brought its unconventional approach to a wider American technology and investment audience.

The company, established in 2024 and based on research conducted at Macquarie University in Sydney, develops conversational AI systems designed to behave like ordinary people who might be vulnerable to scams. Instead of immediately rejecting suspicious calls or messages, the software responds with questions, delays and realistic conversational behavior intended to keep fraudsters engaged while collecting information about their operations.

The 17-year figure represents cumulative conversation time across the company’s network, not 17 calendar years of uninterrupted operation. The figure, along with the reported 2.5 million interactions, comes from Apate and has not been established through a publicly available independent audit. The company considers the time consumed by its agents a measure of disruption, although the more consequential business objective is generating intelligence that its institutional customers can use to prevent fraud.

Apate’s system uses different AI-generated personalities, including variations in age, gender, language and accent. Its agents can answer suspicious telephone calls, respond to text messages and emails, interact with phishing websites and enter online groups where fraudulent investment opportunities or other schemes are promoted. The objective is to create convincing conversations that encourage suspected criminals to expose details about their activities.

Those details can include telephone numbers, bank accounts used to receive fraudulent payments, cryptocurrency wallet addresses, malicious websites and the names of businesses or government organizations being impersonated. Apate processes that information into intelligence products that banks, telecommunications companies and other customers can use to strengthen their defenses.

The strategy originated with an experience involving founder and chief executive Dali Kaafar, a cybersecurity researcher at Macquarie University. During a family picnic, Kaafar received a suspicious telephone call and decided to keep the caller talking instead of ending the conversation. According to an account published by TechCrunch on October 14, the interaction lasted 44 minutes and helped inspire the idea of automating scam-baiting through AI.

Kaafar’s subsequent university research used machine learning and natural language processing to examine the tactics and language patterns found in fraudulent calls. The project eventually developed into a commercial platform. Macquarie University described the underlying technology in 2023, and a university-backed pilot with Commonwealth Bank of Australia was announced in 2024.

The system employs what cybersecurity specialists call a honeypot strategy. Apate and its telecommunications partners operate telephone numbers intended to attract scam calls. When fraudsters contact those numbers, they encounter conversational bots rather than actual consumers. The agents can then prolong the exchange and seek information that might expose a broader campaign.

The platform has also expanded beyond telephone conversations. Apate’s text-based agents can interact across messaging applications and online communities, while its intelligence products organize information about the techniques, infrastructure and organizations associated with suspected scams. Customers can direct agents toward particular investigative objectives, such as identifying accounts or cryptocurrency addresses associated with a fraudulent investment operation.

The technology has moved beyond laboratory testing. Commonwealth Bank confirmed in a 2025 announcement that it was using intelligence generated by Apate’s network of AI agents to improve scam detection and support industry-wide disruption efforts. The bank described the system as a way to generate near-real-time information about emerging criminal tactics.

TechCrunch reported that TPG Telecom had diverted nearly 590,000 calls to Apate’s agents. The startup also reported that an unnamed Australian bank had used its outbound agents to disrupt 22 scam campaigns. These operational figures indicate meaningful deployment, although the extent to which individual interventions prevented losses has not been independently established.

Apate also cited a 76% decline in publicly reported scam losses at Commonwealth Bank. That statistic should not be interpreted as evidence that Apate alone produced the reduction. Financial institutions deploy multiple fraud controls, including transaction monitoring, customer warnings, payment restrictions and information-sharing arrangements. Separating the impact of one technology from those other protections remains an important challenge for evaluating its effectiveness.

The company’s emergence comes as artificial intelligence is also making fraud more sophisticated. Criminals can use generative tools to produce personalized messages, convincing fake identities and synthetic voices. Such capabilities potentially allow fraud operations to approach more targets while reducing the manual work required to conduct each attempted deception.

In the United States, the FBI’s 2025 Internet Crime Report recorded 22,364 complaints involving artificial intelligence, with reported losses of approximately $893 million. The bureau identified applications including fabricated social media identities, voice impersonation and deceptive content designed to imitate trusted individuals or organizations. Those figures describe reported AI-related criminal activity broadly and do not measure the effectiveness of Apate’s technology.

For American banks, payment companies and telecommunications providers, Apate’s approach presents a potential additional layer of protection. Conventional systems often identify suspicious communications, monitor transactions or investigate fraud after a customer reports a problem. Intelligence gathered through direct interactions with suspected fraudsters could provide earlier warning of fraudulent payment destinations, phishing infrastructure and impersonation campaigns.

Such information could be particularly valuable when shared across institutions. A fraudulent bank account identified during one attempted scam may also be associated with additional targets. Similarly, a malicious website uncovered through an AI conversation could potentially be reported or blocked before more consumers are directed toward it. The practical benefits depend on the accuracy, timeliness and responsible handling of that intelligence.

The business model is also attracting venture capital. Apate raised $8.15 million in seed financing in August 2026 from investors including Lobby Capital, OIF Ventures, Investible, Concept Ventures and Baobab Ventures. According to TechCrunch, the company had approximately 20 employees and had raised just under $10 million overall. The latest funding is intended to support product development and international expansion, including a London office and a larger presence in the United States.

Its selection for Startup Battlefield 200 at TechCrunch Disrupt further connected the Australian company with Silicon Valley investors and enterprise technology customers. The event, held at Moscone West in San Francisco through October 15, provided a platform for startups seeking financing, partnerships and broader commercial exposure.

Still, automated scam engagement introduces operational and ethical questions. Institutions adopting such systems must consider how agents distinguish fraudulent contacts from legitimate communications, how collected information is verified and how investigative records are retained or shared. False identification of a telephone number, account or website could create problems if institutions act without adequate verification.

Another challenge is adversarial adaptation. Fraud networks may eventually learn to recognize AI-generated personalities, change their conversational techniques or use automated agents of their own. Long conversations with suspected scammers are not necessarily useful if they produce misleading information or consume resources without generating actionable leads.

Apate’s commercial proposition therefore rests on more than its widely reported 17-year milestone. The lasting measure of success will be whether its intelligence allows customers to identify active scam infrastructure earlier, interrupt fraudulent payment channels and reduce losses without introducing unacceptable privacy or accuracy risks.

For the U.S. technology industry, the Australian startup illustrates a developing category of defensive AI applications: autonomous systems designed not simply to detect malicious activity, but to engage adversaries in controlled environments and extract information. As Apate pursues American expansion, its experience with Australian financial institutions and telecommunications companies could provide an early test of whether that approach can scale across additional markets.