AI Agent Gang Attacks Taiwan: Four-Day Hack of Government Systems

In early July, Taiwanese government websites were attacked by hackers using open-source AI, as reported by The Financial Times citing Israeli AI firm Dream, which first detected the intrusion attempts. According to Dream, the attackers, possibly linked to China, combined several open-source AI agents into an autonomous hacking tool that coordinated their actions. During the four-day attack, the tool used up to eight autonomous agents simultaneously, mapping 21 government systems and searching for vulnerabilities. It constantly reassessed attack priorities based on results; when one method failed, it launched another agent to search the internet and develop a new approach, mimicking a human hacker. Amir Becker, Dream's chief strategist and former head of cyber operations for Israel's elite signals intelligence unit 8200, said he had never seen such a fully autonomous attack on a government target. Dream declined to confirm the specific government targets, citing company policy, but said it had notified a country in the Asia-Pacific region; a Financial Times source identified it as Taiwan. The tool compromised at least 85 government user accounts and accessed over 2,500 personnel records before spreading to Taiwan's nuclear safety agency and at least seven energy companies. The specific AI model used remains unknown, but the defenses were bypassed by framing the hacking actions as an authorized exercise to test system vulnerabilities.