Artificial intelligence: First cyber attack by AI agents on government documented

Security researchers at the Israeli company Dream documented an attack by eight AI agents on the government networks of an Asian country this week. Such an orchestrated autonomous attack on a government target, carried out without the involvement of human hackers, had not been observed by the security researchers before. “We don’t believe governments are prepared for this new threat,” they wrote.

The target country did not want Dream to name it in their report. According to information from the Financial Times, it is Taiwan. For the island nation, this would be a new dimension in an already permanent state of emergency. In 2025, Taiwan's intelligence agency counted 2.6 million Chinese cyberattacks per day, six percent more than in the previous year.

China claims the island and is trying to force the local democracy to surrender through military, diplomatic, and propaganda pressure. Twelve waves of attacks in four days The publicly known case highlights the threat that artificial intelligence can pose to governments, companies, and private individuals. In recent weeks, AI systems from US companies Anthropic and OpenAI have already carried out successful attacks on networks independently - and without the consent of their developers.