India, Oct. 12 -- The risk is visible in recent AI-related incidents, though their causes differ. In July 2026, Hugging Face disclosed an intrusion that began when a malicious dataset exploited code-execution paths in its processing pipeline; attackers subsequently accessed internal datasets and service credentials. In November 2025, Anthropic reported that a suspected state-sponsored group had manipulated Claude Code to automate parts of an espionage campaign.

Neither case can be explained simply as a failed firewall. Both demonstrate how compromised authority and automated capabilities can magnify an intrusion.

Human accounts usually have identifiable owners and established processes for changing or removing access. Service accounts, ...