AI Agents Empower Cybercriminals, Making Scams Cheaper and More Efficient, Experts Warn

Cybersecurity experts are sounding the alarm: AI agents are becoming powerful tools for criminals, enabling them to execute scams with unprecedented efficiency and reduced costs. These AI-powered tools can craft convincing job offers, send messages referencing personal details like workplaces, or even mimic the voice of a relative, all contributing to sophisticated fraudulent schemes.

Criminals are increasingly leveraging AI agents for tasks such as researching potential targets, managing conversations, and executing various stages of cyberattacks. This significantly lowers the barrier to entry, reducing the need for extensive human resources and specialized technical knowledge that was previously required.

The danger of these AI agents lies in their ability to personalize interactions, answer queries, and persistently follow up when initial attempts fail. Farid Zahran, senior managing director of AI and digital transformation at FTI Consulting, noted, “Agents are now doing the work, not just advising.” With the assistance of other tools, AI agents can adeptly gather information, dispatch messages, and maintain contact with potential victims.

The sophistication enabled by AI agents makes it more challenging for victims to detect scams compared to earlier methods that often relied on human error as a telltale sign. Premchand Kurup, chief executive of ParamountAssure, highlighted that this advancement makes certain scams easier and more cost-effective to implement. “Tasks that once required technical expertise, time and a team can now be carried out with fewer resources,” he stated.

However, experts caution that AI is not yet operating autonomously in all scams. Human oversight remains crucial for setting objectives, monitoring attacks, and managing the illicit collection of funds. Scammers utilize AI to meticulously research targets by scouring social media, professional profiles, and company websites before initiating contact. Information about recent holidays, new jobs, or employers can be used to craft believable messages.

“A scam no longer needs to rely on a poorly written message sent to thousands of people. AI can make fraudulent communications more convincing and help criminals tailor their approach to individual targets,” explained Vladislav Tushkanov, group manager at Kaspersky’s AI Technology Research Centre. This allows for tailored approaches, such as a fake recruiter aware of a candidate's work history or a fraudster impersonating a known contact.

While AI excels at processing information rapidly, it does not grant criminals direct access to private messages or bank records, according to Mr. Tushkanov. Unlike static scripts, AI agents can engage in dynamic conversations, remembering previous interactions. If a potential victim questions a payment request, the AI can adapt its strategy or provide reassurances.

“An agent can change its reply based on what someone says and remember what has already been discussed,” Mr. Kurup elaborated. “This can make the exchange feel like a natural conversation rather than a set of fixed replies.” This capability is particularly effective for scams that rely on building trust, such as romance fraud, fake job offers, and investment schemes. AI agents can also manage numerous conversations concurrently, maintaining separate records for each individual.

Nevertheless, human intervention may still be necessary when a target becomes highly suspicious or the conversation deviates unexpectedly. “The ability to generate convincing content doesn’t mean that an agent can reliably impersonate a person or complete every stage of a scam without supervision,” Mr. Tushkanov added.

Attacks are already being documented. A report from Anthropic in August 2020 detailed how an attacker used its Claude Code tool in an extortion operation targeting at least 17 organizations, including healthcare providers, emergency services, and government bodies. The tool aided in network breaches, analysis of stolen data, and the formulation of ransom demands, with some requests exceeding $500,000, as reported by Mr. Zahran. The same report highlighted another instance where an individual with basic coding skills used AI to develop ransomware, with packages offered to other criminals for between $400 and $1,200.

Further cases involving AI assisting in attacks and data theft were described in a September 2026 Anthropic report, where human operators continued to select targets and review stolen information. Microsoft's analysis indicates the UAE is ranked sixth globally as a target for cyberattacks, with the US topping the list in their Digital Defence Report covering January to June of the current year. Other countries frequently targeted include Israel, Ukraine, Taiwan, the UK, and the UAE.

Warning signs for potential scams are evolving. Bad spelling and awkward phrasing are no longer reliable indicators. “AI can write convincing messages that include details about the recipient. A familiar voice or video is not proof of identity either, because both can be faked,” warned Mr. Kurup. Individuals are advised to scrutinize requests closely, looking for unusual demands for money, pressure to act quickly, altered payment details, or requests for sensitive information like passwords and verification codes.

Even if a message appears to originate from a trusted account, verification is recommended. Mr. Tushkanov advises verifying unusual requests through a known telephone number, an official app, or an established contact, stressing, “Avoid using contact details or links supplied in the suspicious message itself.” If a link has been clicked or money transferred, immediate contact with the bank or payment provider is crucial to attempt stopping or recalling the payment, followed by reporting the incident to the police. Preserving evidence such as messages, screenshots, and payment references is important. If a password has been compromised, it should be changed on a trusted device, and any other accounts using the same password should be updated. Those who have downloaded suspicious files or installed software should seek technical assistance, especially if unauthorized device access is suspected.

Mr. Tushkanov cautioned against continuing the conversation in an attempt to recover lost funds, stating, “Early action may improve the chances of limiting further losses, although recovery cannot be guaranteed.”