Runtime security layer for AI agents that analyses full execution sequences, including chained tool calls, retrievals and agent-to-agent handoffs, to detect direct and indirect prompt injection, unauthorised tool invocations and sensitive data leaving through agent activity, with inline blocking. Evidence is drawn from vendor platform pages.
commercial · generally available · Research snapshot 2026-09-06
Show me AIDR detecting a slow-building indirect injection across several turns, blocking the unauthorised tool call it triggers, and the agent-to-agent handoff record.
Capabilities and evidence
Support labels reflect the supplied research. Documentation and vendor claims are not independent product tests. “Not established” means the researcher did not find support; it does not prove a capability is absent.
Vendor claim
Zenity states AIDR reads the full sequence of chained tool calls, retrievals and actions, detects direct and indirect prompt injection including paraphrased and slow-building attacks, and blocks execution before the agent acts.
Limit: No reference documentation, detection benchmark or false-positive guidance was fetched.
AIDR is described as flagging and blocking unauthorised tool invocations or API calls that violate policy and monitoring and blocking sensitive data leaving through agent conversations, tool calls or encoded payloads in real time.
Limit: Policy authoring model and supported enforcement points are not documented on the page.
Zenity's documentation site lists Zenity Observe, Zenity Govern and Zenity Defend covering observability, AI security posture management and AI detection and response.
Limit: The docs landing page does not detail runtime enforcement mechanics.