security research (Acronis Threat Research Unit report, Nov 4), press (eSecurity Planet)
One hacker used 12 AI agents to target groups in 11 countries, Acronis researchers say
The attacks ran on local models after commercial ones refused.
Key facts
- Who
- A single operator, assessed with high confidence to be based in Hualien City, Taiwan (Acronis)
- Tools
- A mesh of 12 AI agents, nine of them active, linked to open source hacking tools (Acronis)
- Reach
- Targets in 11 countries over 12 days, including emergency services, banks, universities and military systems (Acronis)
- Breaches
- Data theft or system access confirmed at roughly 10 targets in three countries: Taiwan, Mexico and Thailand (Acronis)
- Refusals
- Commercial models refused offensive tasks, so attacks ran on local models with safety training removed (Acronis)
Every prompt aimed at vulnerability identification or exploitation was met with resistance.
Acronis Threat Research Unit, on commercial models
Note
A single operator used a mesh of 12 AI agents, called BOSSAI, to target organizations in 11 countries, the Acronis Threat Research Unit said in a report on Wednesday, Nov 4.
Taiwan was hit hardest. Agents pulled emergency dispatch case records and reached exposed patient care data at a Taipei fire department platform, Acronis says. Confirmed data theft or system access occurred at roughly 10 targets in three countries.
Commercial AI models used to build the system refused to help with attacks, the researchers say. The operator turned to locally run models with safety training removed, which were less capable and often invented findings.
Acronis says it assesses with high confidence that the operator is an individual based in Hualien City, Taiwan. No arrest has been reported.
Why it matters
It shows one person running a multi agent attack system, and that safety training in commercial models held while stripped local models did not, Acronis says.
Sources
- Acronis TRU: BOSSAI, inside the AI attack mesh that targeted Taiwan's emergency services (Nov 4 2026)
- eSecurity Planet: BOSSAI AI agents target Taiwan emergency services (Nov 2026)
Primary + secondary attribution as listed.
Doom Signals desk · About · Contact · Corrections. Spotted an error? Write the desk; fixes go on the public log.