How to meet it
OWASP AISVS v1.01
Input Validation · Prompt Injection Defenses
Catch encoding tricks hidden in input
Verify that encoding and representation smuggling in inputs is detected and mitigated. Approved mitigations include canonicalization, strict schema validation, policy-based rejection, or explicit marking.
OWASP AISVS v1.01, requirement 2.1.2
In plain English
Catch instructions hidden with encodings and tricks, such as Base64, homoglyphs, invisible characters or text hidden in markup, before they reach the model.
What good looks like
- Known encodings are detected and either decoded and re-checked or rejected.
- Structured input is validated against a strict schema.
- Hidden content in documents, such as white text or comments, is stripped or flagged.
How teams usually do it
- Decoding common encodings and running the decoded text through the same injection checks.
- Document extraction that drops hidden layers, comments and metadata.
Evidence an auditor would accept
- Test cases covering encoded and hidden instructions.
- The extraction and decoding configuration.
Common mistakes
- Checking only the visible text of uploaded documents.
- Decoding once but missing nested encodings.
Not sure which requirements apply to your system? Take the assessment.
Requirement text from the OWASP AI Security Verification Standard, licensed under CC BY-SA 4.0. This guidance is shared under the same licence.