Acronis reports 12-agent campaign targeting emergency services

Acronis says investigators discovered the campaign while examining what appeared to be a misconfigured development environment.[1] The exposed environment contained more than 90,000 files and about 9GB of evidence, including logs, agent conversations, tools and victim data.[1] Acronis says one operator used the BOSSAI mesh in campaigns across several countries and targeted hundreds of organizations in sectors including emergency services, finance, education, government and the military.[1] The discovery provided investigators with material to examine how the operator coordinated agents and what targets they reached.[1]

The BOSSAI mesh contained 12 agents, nine of which were active when researchers examined it, according to Acronis.[1] The operator manually passed information between agents rather than enabling them to communicate directly.[1] Acronis says builder agents maintained infrastructure while attack agents scanned targets and developed exploits.[1] The agents used MCP-connected tools including HexStrike AI, a Kali Linux server and a Ghidra bridge.[1] CyberSecurityNews reports that Hermes was the most active attack agent, with about 95 sessions and 18,311 messages.[2] These details describe a coordinated workflow, but not a fully autonomous agent-to-agent operation.[1]

CyberSecurityNews reports that a Taipei fire-department patient-care platform exposed ECG readings, ambulance video and vital signs from 2018 without requiring a login.[2] Acronis says agents also retrieved case records from a city 119 dispatch system covering seven districts.[2] In confirmed breaches at universities and emergency-services providers, Acronis says stolen information included medical data and personally identifiable information.[1] The reported activity also extended beyond Taiwan: Acronis says agents collected about 206,800 residential phone records in Mexico, while CyberSecurityNews reports that agents identified 489 targets across 45 universities in Thailand and retrieved data from three universities.[2]

The campaign’s reported scale does not mean its agents were consistently effective.[1][2] Acronis says prompts to identify or exploit vulnerabilities met with resistance, after which the operator turned to locally run abliterated models that were less effective and resource-constrained.[1] More than half of the campaign’s high-severity findings were false positives, according to CyberSecurityNews.[2] Acronis also says agents passed over potential lucrative targets, including a Honduran military base, because they deemed them out of scope.[1] For defenders, Acronis recommends patching exposed databases, removing default credentials, closing old ports and requiring login checks on every API, including those serving dispatch and medical data.[2]