CrowdStrike identified four previously unknown threat groups that used agentic artificial intelligence to conduct intrusions with limited human involvement. The agents performed reconnaissance, exploited vulnerabilities and stole data, targeting healthcare organizations, government agencies and software-as-a-service providers across multiple countries. One campaign compromised at least 15 healthcare organizations in Argentina and stole medical records and credentials, while another agent completed lateral movement and domain credential harvesting in about 20 minutes without human guidance. According to CrowdStrike’s telemetry, agentic intrusions tripled in the second half of the year compared with the first half, and their volume doubled between July and August 2026. Falling inference costs and improving model capabilities could make these attacks accessible to less skilled operators, shorten defenders’ response window and increase the exposure of the healthcare sector.
Source: Read the original article
