Control Efficacy

Control efficacy refers to how well a security, compliance, or governance control actually does the job it was designed to do, as opposed to whether the control merely exists on paper. In AI governance, it asks whether a policy, technical safeguard, or audit procedure reliably prevents or detects the harm it targets, such as biased outputs, data leakage, or unauthorized model changes. Regulators and auditors increasingly look past documented controls to test whether they work in practice, so organizations need evidence of effectiveness, not just evidence of intent.

1 item

ResearchUS2026-11-01

80% of Agent Incidents Lack Control Evidence, Anaconda Finds

Anaconda, joining the Open Secure AI Alliance, published research finding that 80% of recorded AI agent incidents lacked public evidence that security controls existed or were operating. The research also argues that testing individual prompts or tool calls is insufficient because retrieved content can silently alter an agent's plan and trigger unintended actions mid-task. Both findings point to a documentation gap that regulators and auditors are increasingly positioned to exploit.

Anaconda red teaming agentic AI incident reporting prompt injection control efficacy audit evidence