Analysis
Indirect Prompt Injection: Protection at Runtime
Publication date unknown · Discovered fiddler.ai
Page published
Publication date unknown · First observed: 18 Oct 2026
Coverage timeline
Single-source analysis — one report is available.
Why it matters
Indirect prompt injection can combine an agent's private-data access with outbound capabilities to enable exfiltration, making capability restrictions and traceability important defensive controls.
Fiddler's explainer describes indirect prompt injection through retrieved content and tool outputs, citing EchoLeak (CVE-2025-32711) in Microsoft 365 Copilot as a historical example of sensitive-data exfiltration. The article recommends treating external content as untrusted, enforcing capability limits in deterministic code, and tracing retrievals and tool calls rather than relying on input filtering alone.