As enterprises rush to deploy AI agents to handle their workflows, they’ve accidentally built a blind spot for themselves. New data from ExtraHop reveals that AI has created a highly complex attack surface for enterprises to defend. 83% of UK organisations battled AI-driven security incidents, data exposures or "near-misses" in the last 12 months, with traditional security tools failing to keep pace with autonomous, AI traffic. More than half of UK IT and security leaders now point directly to AI agents, agentic infrastructure and Generative AI as their single most significant cybersecurity risk. Only 6% of UK organisations can say with absolute certainty that they did not experience an AI-related security incident or data exposure last year.