Disclosed vulnerability
Amp Code interpreted invisible Unicode Tag characters as instructions
AIR-2025-0100 AIR record · 2025-08-amp-code-invisible-injection
From the supporting source
Amp was vulnerable to interpreting invisible Unicode Tag characters as instructions.embracethered.com
The record
Disclosed
2025-08-16
Vendor
Sourcegraph
Product
Amp Code
Risk domain
Privacy › Data exfiltration channels
System type
Generative model
Who acted
Agent acted — Amp Code invoked shell tools from invisible instructions
Attack surface
Coding agent
Autonomy level
Semi-autonomous
Initial vector
Invisible unicode injection
Impact type
Data exfiltration
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI02 – Tool Misuse and Exploitation
- MITRE ATLAS
- AML.T0024 – Exfiltration via AI Inference API
- NIST AI RMF
- MAP 4.2
- EU AI Act
- Article 10
- ISO/IEC
- TR 24028; 23894
- AIUC-1
- A. Data & Privacy
- Scope
- tool misuse
- Mitigation
- Monitor network traffic for anomalous patterns; Restrict egress destinations.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
Amp Code users
Capability abused
Defense & remediation
Recorded defense
human review
Guardrail kinds
Unknown
Guardrail outcome
Silent
Fix shipped
fixed - input sanitization added shortly after the 2025-06-14 report
Taxonomy & references
Curator notes
Month of AI Bugs 2025.
Supporting sources
Cite as
Enkrypt AI. "Amp Code interpreted invisible Unicode Tag characters as instructions". AIR, AIR-2025-0100. https://agentincident.vercel.app/i/AIR-2025-0100/ (accessed 2026-10-19).