Threat researchCisent
An attacker used AI agents to break into at least 15 healthcare organizations in Argentina
CrowdStrike found groups that hand their intrusions over to AI agents, and one of them went after Argentina's healthcare sector, taking passwords and medical records.

On October 21, 2026, CrowdStrike’s team that tracks attackers published four campaigns in which previously unknown groups handed their intrusions over to AI agents. One of them compromised at least 15 healthcare organizations in Argentina.1
What happened
All four groups made the same mistake: they left the directories holding their agent tooling exposed on the internet. That’s how CrowdStrike was able to reconstruct how they worked.1
The group that went after Argentina’s healthcare sector speaks Spanish and English. It used an open-source penetration-testing tool, modified so that AI agents could run it, and CrowdStrike saw it operating from mid-May to mid-August. It stole passwords and, at at least one of the victims, medical records.12
Its agents jumped from one victim’s network to another’s and switched techniques on the fly. That’s how they took five passwords from one organization’s domain controller, the server that manages access across its whole network.2 CrowdStrike doesn’t name the organizations affected.1
The other three groups speak Chinese, Indonesian and Russian, and they attacked governments and companies in Brazil, Colombia, the United States, Indonesia and other countries. The Russian-speaking group gave its agent a seven-item to-do list and near-total autonomy, and took customer data from a U.S. help-desk platform.2
Why it matters
Two of the four groups used simple tooling and still got in. According to CrowdStrike, what holds back an AI-agent campaign the most isn’t engineering, but how much the attacker can spend on running the models.2
“This delegation allows AI to operate at potentially faster speeds, and for longer durations, than some human actors,” CrowdStrike wrote in its report.2 And it expects these intrusions to grow in volume and sophistication.1
Where Cisent comes in
Sentinel Red Team tests the organization’s systems the way an attacker would, with its authorization and within the scope it defines: what’s exposed, how it’s configured and where someone could get in, without exploiting anything or affecting availability. If attackers already have agents that test around the clock, testing can’t happen once a year: Red Team repeats it every day, week or month, and every gap it finds becomes a Sentinel Defense rule.
Sources
CrowdStrike Counter Adversary Operations, “CrowdStrike Uncovers New Agentic AI Campaigns”, October 21, 2026. https://www.crowdstrike.com/en-us/blog/crowdstrike-uncovers-new-agentic-ai-campaigns/↩︎
Eric Geller, Cybersecurity Dive, “Agentic AI helps hackers level up, even without detailed instructions”, October 22, 2026. https://www.cybersecuritydive.com/news/agentic-ai-cyberattacks-harnesses-crowdstrike/833714/↩︎