Social Engineering and AI Cybercrime: 2026 Reports

Key Takeaways

• The supplied sources do not substantiate a Google DeepMind–GovTech Singapore on-device scam-detection project or the draft’s associated Singapore statistics.
• The National reports that a person posing as an Italian official reportedly obtained private information on about 650 Revolut users. Experts characterized the method as social engineering.
• The same report says there was no indication that AI was used in the Revolut incident. An expert said AI is making similar attacks cheaper, faster, and easier to scale.
• The Wall Street Journal reports a separate attack in which hackers used a Chinese AI agent against South Korean banks and stole personal information belonging to 68,000 people.
• A separate WSJ report says a processor rejected more than 80% of the deposits it handled for Polymarket US as fraudulent at one point. This is not an estimate of social-engineering losses.
• The National reports 33 arrests in an international operation involving an alleged $2.4 billion money-laundering scheme, including a group accused of using social engineering to steal digital wallets.

Revolut Report Highlights the Human Side of Cybercrime

A September 2026 report by The National describes an incident in which a person posing as an Italian official reportedly requested and received personal information associated with about 650 Revolut users. The publication reports that the attacker used legitimate government email addresses to socially engineer the disclosure. The reported incident is a reminder that an organization can face risk not only from technical vulnerabilities but also from people being persuaded that a request is legitimate.

The article quotes cybersecurity experts describing the activity as “ClickFix-style social engineering.” In the explanation quoted by The National, a victim is persuaded to run malware under the pretext of installing software or fixing a technical issue. The source also describes screenshots and subsequent harassment as tactics apparently intended to pressure a target into setting aside doubts. These details are reported observations and expert characterizations; they should not be treated as a complete forensic account of every step in the incident.

The distinction between social engineering and a technical exploit matters. A deceptive request can exploit trust in an official-looking sender, a familiar process, or a claimed technical problem. The attack may ask a person to disclose information or take an action rather than directly defeat an organization’s software defenses. The source does not establish that every tactic described in its reporting was used in the Revolut incident, so it is important not to merge the article’s examples into a single unsupported technical narrative.

The article also reports that technology, culture, and philosophy writer Joan Westenberg described a separate attempted intrusion disguised as a podcast invitation. According to her account, the people behind the approach supplied convincing interview questions and then offered a link that asked her to run a terminal command to install webinar software. She says she recognized the risk and stopped communicating; she also reported continued pressure and harassment afterward. Her advice, as quoted in the report, was not to install custom software for a meeting or podcast.

That example is a reported personal account, not evidence that the same people or method were involved in the Revolut case. It does, however, illustrate why requests to install unfamiliar software or execute commands deserve scrutiny. A meeting invitation, interview, or supposed technical fix can be used as a pretext. The safe response depends on verifying the request through an independent channel rather than relying only on the contact details or links provided by the person making it.

What the Sources Say About AI and Attacks

The reporting makes two different points that should not be conflated. First, The National says there is no indication that AI was used in the particular Revolut incident. Second, an expert quoted in that article says the growing availability of AI is making similar attacks cheaper, faster, and easier to scale. The first is a statement about the evidence for one reported case; the second is an expert’s broader assessment of how AI may affect comparable activity.

The Wall Street Journal separately reports that hackers used a Chinese artificial-intelligence agent to attack South Korea’s biggest banks and steal personal information belonging to 68,000 people. The summary available here calls it one of the first such AI-powered intrusions into the global financial system, according to the Journal. The provided excerpt does not identify the agent, describe its architecture, explain the precise actions it performed, or provide further technical details. It would therefore be premature to claim from this excerpt that the agent autonomously completed a particular stage of the attack or that the incident used a specific model or exploit.

Taken together, these reports support a measured conclusion: AI is part of the current discussion about cybercrime, and one report describes an attack involving an AI agent, while another cautions that AI may increase the speed and scale of similar social-engineering campaigns. They do not establish that AI was involved in every fraud or intrusion reported here. Clear attribution matters because overstating...