Agentic AI scales semiautonomous server attacks

Cisco Talos researchers identified a Chinese-speaking cybercriminal group dubbed UAT-10147 that uses agentic AI to scale and automate offensive operations—targeting government, media, technology, and gaming sectors—by exploiting publicly disclosed vulnerabilities, deploying custom tooling (SPECTRE backdoor, Specter rootkit), and automating post-compromise activities and playbooks, driving the recommendation that SOCs adopt agentic AI capabilities to keep pace.