Dream says suspected China-linked hackers launched autonomous AI attack on Taiwan, stole 2,500 records
This digest was compiled by AI from multiple sources — links to the originals are below.
Israeli cybersecurity firm Dream said on August 12 that hackers with suspected links to China used publicly available AI tools to carry out an autonomous cyberattack on Taiwanese government systems, compromising at least 85 accounts and stealing more than 2,500 personnel records. The campaign reportedly lasted four days in early July, deploying as many as eight autonomous agents in parallel, and affected Taiwan's nuclear safety agency and at least seven energy companies. Dream stopped short of attributing the campaign to a specific country, but said internal operator communications were written in Simplified Chinese.
The AI-Agent Toolkit
Dream said the operators assembled an autonomous hacking platform from two open-source AI-agent frameworks, Hermes and OpenClaw, both freely downloadable and designed to let large language models run multi-step tasks. The platform mapped 21 government systems and at times ran eight agents in parallel to map networks, research vulnerabilities, attempt intrusions, and adapt after failed attempts. Researchers found evidence in a 160-megabyte online archive containing 1,395 files, according to an August 12 Financial Times report. The underlying model was not identified, but Dream said safeguards were sidestepped by presenting the intrusion as an authorized penetration test.
Taiwan Government Systems Breached
The campaign compromised at least 85 user accounts and exfiltrated more than 2,500 personnel records from Taiwanese government systems during four days in early July, Dream said. The attackers later expanded to Taiwan's nuclear safety agency, at least seven energy companies, government suppliers, and other systems. Dream said it notified a country in the Asia-Pacific region but declined to name the victim, citing policy. The tool's operators reprioritized attack paths as circumstances changed, with failed techniques triggering new internet searches for alternative approaches.
Attribution Evidence
Dream stopped short of attributing the campaign to a specific hacking group or country, but said internal operator communications were written in Simplified Chinese, indicating a high probability of a China connection. The company's researchers describe the operation as the first observed end-to-end autonomous cyberattack against a government target. The toolkit used no purpose-built offensive software, drawing concern from researchers about easily available components.
What's Next
The notified Asia-Pacific government is expected to conduct its own forensic review and may issue a public alert. It remains unclear whether Dream's evidence will lead to formal attribution or policy changes for open-source AI-agent tools.
5 sources
Dream says suspected China-linked hackers launched autonomous AI attack on Taiwan, stole 2,500 records
