13/08/2026
Share:
13/08/2026
Share:
TAIPEI, Aug 13: Hackers used an autonomous artificial intelligence system to conduct a sophisticated cyberattack against government agencies and critical-sector organizations in Taiwan, officials and cybersecurity experts said, in what they described as a potentially unprecedented case of AI-driven attacks operating with limited human intervention.
According to Israeli AI security firm Dream, the campaign took place over four days in July. The attackers reportedly mapped 21 government systems, compromised 85 government user accounts and extracted about 2,500 personnel records. The operation also targeted Taiwan’s nuclear safety agency, government IT suppliers and at least seven companies in the energy sector.
Taiwan’s Ministry of Digital Affairs said the investigation found clear indications that the attacks originated overseas and employed a hybrid approach that combined conventional hacking techniques with AI agents. The ministry did not publicly identify the perpetrators.
Experts have suggested that the operation may have links to actors based overseas, but Taiwanese authorities and Dream have not confirmed the attribution. The use of distinctive character forms in internal documents associated with the intrusion was cited as one indication, although this alone does not establish responsibility.
Dream said the AI system coordinated as many as eight AI agents, which carried out reconnaissance, credential attacks, exploitation of vulnerabilities, and planning of subsequent attack paths. Unlike conventional AI-assisted hacking, the system reportedly adapted its tactics when blocked and continued operations without requiring a human operator to direct every step.
Kenny Huang, chairman of the Taiwan Network Information Center, described the incident as the first fully automated cyberattack against a government he was aware of. The incident has renewed concerns that increasingly capable AI systems could lower the cost and increase the scale and speed of cyberattacks.
Taiwanese officials said the episode demonstrated the need for stronger defensive capabilities, legal frameworks and policies to address AI-enabled cyber threats, as governments worldwide confront the growing challenge of autonomous systems being used for malicious purposes.
Meanwhile, according to AP, For 30 minutes, the internet in parts of Taiwan crawled.
The slowdown Thursday for people in the capital, Taipei, and six other cities and localities was part of the last phase of the island’s annual Han Kuang exercises, military drills that began last week and aim to model a response to a possible Chinese military invasion.
Telecommunication operators slowed mobile web traffic to 256 Kbps, making it difficult to exchange anything other than texts. Usually, residents can access 5G in major cities.
Taiwan’s President Lai Ching-te apologized to the public for the inconvenience: “We ask for your understanding and support, as this is a necessary drill,” he said.
Taiwan has stepped up the intensity of the Han Kuang drills, incorporating reservists and more civilians into the exercises with a focus on real-world scenarios. Previously, urban resilience drills in Taipei went largely unnoticed as they were not disruptive. For example, air raid drills in the metro station didn’t stem the flow of passengers. This year, integrating civilians and coordinating with local governments was a key part of the exercises, according to the Taipei Times. At a Taipei metro station Thursday afternoon, passengers waited to be released after the station’s gates were shut for the simulation as part of air raid drills.
