Analysis
Invisible Unicode Prompt Injection: Runtime Controls
Publication date unknown · Discovered trussed.ai
Page published
Publication date unknown · First observed: 15 Nov 2026
Coverage timeline
Single-source analysis — one report is available.
Why it matters
Invisible-Unicode prompt injection can evade visual review, making layered input handling and external tool-call enforcement important safeguards against unauthorized agent actions.
Trussed AI's guide explains how Unicode tag characters, zero-width characters, and bidirectional controls can conceal prompt-injection instructions in content consumed by AI agents. Recommended defenses combine input normalization and filtering with externally enforced tool-call policies, least-privilege permissions, approval workflows, and audit logging.