Untrusted input
Untrusted input is externally supplied data or content whose safety, integrity, or reliability has not been established for its processing context.
AI Governance
Untrusted content is externally supplied information that has not been verified as safe or reliable.
- How it is used
- Frames externally supplied content as a security risk factor when assessing GenAI applications.
- How deep it goes
- Defines prompt injection and jailbreaks, then traces how data access and communication can escalate consequences.
- Examples
- Uses adversarial chatbot prompts, hidden instructions in RAG documents, image-based injection, and browser agents as examples.
- What you can do after
- Evaluate how untrusted input combines with data access and external communication to create data-leak paths.
Prompt Engineering in Practice
Externally supplied material that must not be treated as authoritative instructions without additional controls.
- How it is used
- Treats untrusted input as a prompt-security risk to manage when runtime text enters production prompts.
- How deep it goes
- Defines trust by control and explains how unbounded prompts let runtime text influence earlier instructions, but does not detail defenses.
- Examples
- Uses a request to override instructions and expose customer records, alongside user messages, retrieved documents, and API responses.
- What you can do after
- Evaluate how external text can influence prompt behavior and recognize the need to separate it from trusted instructions.
Read how Prompt Engineering in Practice explains untrusted input →
Also covered in
- Grokking Web Application Security
- GitHub Actions in Action: Continuous integration and delivery for DevOps
- LLM Evaluation and Alignment, The Foundational IdeasIn early access (MEAP)
- Context EngineeringIn early access (MEAP)
- Vibe EngineeringIn early access (MEAP)
- Software Security for Developers
- Vanilla WebIn early access (MEAP)
What to know first
- language modelA language model is a computational model that represents patterns in linguistic sequences to assign probabilities, predict, generate, or evaluate language.
- authorizationAuthorization is the determination of which actions or resources an identity or other entity is permitted to access under defined permissions or policies.
- guardrailA guardrail is a constraint, control, or monitoring mechanism that limits system behavior to reduce risks and keep actions or outputs within acceptable bounds.
- promptA prompt is text or other input supplied to an AI model to provide instructions, context, or other information that guides its response or generated output.
Related concepts
- trusted inputcontrasts with · Trusted input is data considered sufficiently reliable or authoritative within a given trust model to use without additional validation.
- Lethal Trifectapart of · The Lethal Trifecta is a security-risk pattern combining sensitive data access, untrusted content exposure, and external communication capability in an AI system.
- authorization decisionuses · An authorization decision determines whether a principal is permitted to perform an action or access a resource under applicable rules or policies.
- delimiter-based separationused by · Delimiter-based separation is a prompt-hardening technique that marks the boundaries of untrusted content so a model is less likely to interpret that content as instructions.
- evaluses · eval is a language operation or built-in function that interprets and executes a code representation within a specified execution context.
- prompt injectionused by · Prompt injection is an attack that introduces instructions through input or context to manipulate an AI system’s behavior or override its intended instructions.
Questions about untrusted input
What is untrusted input?
Untrusted input is externally supplied data or content whose safety, integrity, or reliability has not been established for its processing context.
Which Manning books explain untrusted input?
AI Governance and Prompt Engineering in Practice.
What should I know before learning untrusted input?
Manning books that teach untrusted input usually introduce language model, authorization, guardrail and prompt first.
What is the difference between untrusted input and trusted input?
Untrusted input is externally supplied data or content whose safety, integrity, or reliability has not been established for its processing context. Trusted input is data considered sufficiently reliable or authoritative within a given trust model to use without additional validation.
What is untrusted input part of?
Untrusted input is part of Lethal Trifecta. The Lethal Trifecta is a security-risk pattern combining sensitive data access, untrusted content exposure, and external communication capability in an AI system.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.

