Your AI agent could become your biggest insider threat

DTEX researchers found that agentic AI tools like Anthropic’s Claude Cowork, when used in corporate environments, can be exploited for rapid data exfiltration through features like the Dispatch tool. Scenarios tested showed sensitive data could be stolen within 10-30 minutes, raising concerns about insider threats and nation-state actors using AI to bypass security. North Korean operatives working in Western tech firms were highlighted as a specific risk.