AI agents used to steal Taipei emergency dispatch records

A single attacker used AI agents to extract emergency dispatch records covering seven Taipei districts, according to researchers at Acronis's Threat Research Unit, whose findings were reported on 9 November 2026. The same agents also turned up a Taipei fire department platform that exposed patient data from 2018, including ECG readings and ambulance video footage, with no login required.

The tooling involved is described by the researchers as BOSSAI, a set of AI agents. In this case the agents were doing the work that a human intruder would normally do by hand: browsing systems, finding what was reachable and pulling data out. That is the detail that makes the case notable rather than routine. The volume of material taken, and the breadth of it across seven districts, came from one person directing software rather than from a team.

Acronis's researchers said they are highly confident that the attacker is a single person. They did not, in the material reported, attribute the activity to any group or state, and no motive has been established. The Agentic Times is not drawing any conclusion about who the person is.

The second finding is arguably the more uncomfortable one for Taipei. A fire department platform was reachable without authentication and was serving patient data from 2018. Emergency medical records are among the most sensitive categories of personal data a city holds. ECG readings identify a specific patient's cardiac activity at a specific moment, and ambulance video footage can show patients, bystanders and locations. None of that required a password to reach, according to the researchers.

The case is a practical illustration of a problem that security teams have been warning about for roughly two years: AI agents lower the cost of reconnaissance. Scanning for unsecured systems, following leads and collecting data used to be slow, manual work that limited how much ground one attacker could cover. Agents remove that limit. An exposure that might have gone unnoticed for years because nobody had bothered to look can now be found by software that never stops looking.

No vendor has been named as responsible for a flaw here, and no software defect has been disclosed. The root problem described by the researchers is a system left open and a set of general-purpose agents pointed at it. The incident is currently at the reported stage. It is not clear from the available material whether the exposed fire department platform has since been secured, or whether affected patients have been notified.