48% of AI Agents Lack Security Controls, Exposing Enterprises to Risks

• •48% of AI agents in enterprises lack meaningful security controls.
• •88% of organizations with deployed agents have experienced security incidents.
• •The average cost of an AI agent-related data breach is $4.7 million.

A report by NeuralTrust reveals that 48% of AI agents in enterprise environments operate without meaningful security controls, posing significant risks. The study highlights that 88% of organizations with deployed agents have faced at least one security incident, with 34% of agents affected by prompt injection attacks. The average cost of an AI agent-related data breach is estimated at $4.7 million. As AI agent deployment doubled between December 2025 and April 2026, monitoring coverage remained stagnant at 52%, widening the security gap. Traditional application security measures are insufficient for these systems, which behave differently than standard applications. The report identifies ten critical threat categories, emphasizing the need for new security strategies tailored to AI agents.