Suspected Chinese cyber operatives used publicly available AI tools to compromise Taiwanese government systems before expanding the campaign to Taiwan’s nuclear safety agency, supply-chain vendors, and at least seven energy companies, according to The Register. The report is based on research from Dream, an Israeli cybersecurity firm, which described the incident as a “near-autonomous attack.” Dream said the activity unfolded over the first four days of July, compromised 85 government user accounts, and extracted more than 2,500 personnel records. Researchers found evidence in a 160 MB online archive containing 1,395 files documenting the operation, The Register reported. Dream’s public research referred to targets as “government entities in Asia” and did not name the government. The Register said a person familiar with the attack confirmed Taiwan was the target, and noted that the Financial Times had first reported Dream’s research and identified Taiwan. The attribution remains deliberately narrow. Dream did not attribute the operation to the Chinese government or to a specific hacking group, according to The Register. The firm said the operational documentation pointed to a Chinese-language operator, which supports suspicion about the operator’s language environment but does not by itself establish state direction. Technically, Dream said the attack framework was built on the open-source Hermes and OpenClaw AI agents. The system deployed up to eight sub-agents, each assigned targets and attack techniques, across 12 attack waves between July 1 and July 4, according to The Register’s account of the research. The first phase was reconnaissance. The agents reportedly mapped a government ecosystem by extracting embedded URLs, application programming interface endpoints, OAuth client IDs, and Keycloak configuration objects from a single government portal. From that portal, Dream said the agents identified 21 connected government systems and every supported authentication flow. The findings described a familiar enterprise weakness at machine speed: exposed internal interfaces and brittle identity controls. Dream said one target alone exposed more than 36 API endpoints across account management, user-data retrieval, file upload, and administrative functions, many of which were unauthenticated. The researchers also said one system exposed its full user database without authentication, including employee names, departments, and single sign-on account IDs. After mapping the environment, the agents found several entry points, including three hidden API endpoints that returned valid authenticated sessions without requiring user credentials, according to The Register. Using employee usernames harvested from an unauthenticated API, the agents then attacked a government department’s office automation portal, reportedly solving CAPTCHAs with 100% accuracy and testing predictable password patterns based on employee IDs. Dream said the agents cracked 85 accounts over multiple password-spray rounds, and 84 of those accounts successfully authenticated to the department’s internal information system. That access exposed internal dashboards, equipment management interfaces, personnel statistics pages, more than 2,564 personnel records, a full JSON export of department system users, seven single sign-on client secrets, and six internal database credentials across systems including MSSQL and Oracle, according to The Register. Who benefits: Security teams that maintain strict API inventories, test authentication flows, and monitor password spraying are better positioned against this style of campaign. Vendors offering API security, identity hardening, and automated exposure management may see stronger demand if similar incidents are corroborated elsewhere. Who's exposed: Government agencies and critical-infrastructure operators with unauthenticated internal endpoints, weak single sign-on controls, or predictable password conventions are most exposed. Supply-chain vendors connected to sensitive public-sector environments also face risk if attackers use government access to pivot outward.