Unicode/Encoding Bypass
Use Unicode homoglyphs, invisible characters, or encoding tricks to bypass input filters and inject instructions
Initial Access · Stage 2
Gain control over agent behavior through prompt manipulation or input exploitation
UNICODE-STEGO
Using invisible Unicode characters, homoglyphs, and encoding tricks to bypass filters
Confirmed in real-world production systems or internet-wide exposure assessments.
token-smuggling-unicode
Reproductions in Damn Vulnerable AI Agent, the OpenA2A intentionally-broken agent for kill-chain validation.
Recent observations
Technique-attributed observations from the AgentPwn + TrapMyAgent fleet, classified into Threat Matrix techniques by the registry evidence bridge. See the Attack Prevalence Index for the full distribution.
AgentPwn coverage
An AgentPwn trap page produces a payload tagged with this technique class. Following the AgentPwn taxonomy of trap pages shows what an agent encounters.
Higher tiers carry base64, Unicode homoglyph, and zero-width payloads.
Evidence by source
Evidence timeline
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-007 on agentpwn.com (prompt-injection tier 7; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-007 on agentpwn.com (prompt-injection tier 7; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-007 on agentpwn.com (prompt-injection tier 7; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-JB-005 on agentpwn.com (jailbreak tier 5; classified browser/claimed)
Client fetched the canary embedded in APWN-PI-008 on agentpwn.com (prompt-injection tier 8; classified browser/claimed)
Detection · HackMyAgent
npx hackmyagent secure --ciLive = implemented in hackmyagent; queued = declaredDefense · OASB controls
How to cite
AI Agent Threat Matrix T-2006 (Unicode/Encoding Bypass). OpenA2A, 2026. https://threats.opena2a.org/techniques/T-2006