BOSSAI: Inside the AI attack mesh that targeted Taiwan’s emergency services

Acronis researchers have described BOSSAI as an interconnected mesh of open-source and “abliterated” AI models alongside commercial and closed-source systems. According to the write-up, the models are assigned specialised offensive functions and are supported by MCP servers. Acronis describes the resulting platform as capable of highly automated sophisticated attacks. The article’s headline states that the mesh targeted Taiwan’s emergency services. The write-up does not name the commercial or closed-source systems involved.

The write-up states that the open-source agent CyberStrike appears to have run vulnerability assessments and exploitation attempts against live targets. CyberStrike is described on its GitHub page as the first open-source AI agent built for offensive security. Acronis says the agent’s specific outputs could not be independently verified. The report does not state whether the exploitation attempts succeeded, nor does it identify who operated the mesh.

The write-up frames BOSSAI as a platform capable of highly automated sophisticated attacks. It identifies MCP servers as supporting infrastructure for the mesh. It also describes CyberStrike as having been used in vulnerability assessments and exploitation attempts against live targets, while noting that the specific outputs could not be independently verified. The headline identifies Taiwan’s emergency services as the target of the activity described. The write-up does not provide a date for the activity or a count of the models in the mesh.