# Input Filtering: definition and example

URL: https://airuntimesecuritytools.com/glossary/input-filtering/
Updated: 2026-11-16

Input filtering inspects what is sent to a model, including user messages and retrieved documents, and can block or change it before the model sees it. It is the usual place to catch injected instructions, jailbreak attempts and sensitive data in prompts. A filter on text alone cannot see what an agent does afterward.

## Example

Cloudflare says AI Gateway Guardrails work as a proxy between the application and model providers such as OpenAI and Anthropic.

## Sources

- [Cloudflare page](https://developers.cloudflare.com/ai-gateway/features/guardrails/) (developers.cloudflare.com), read 2026-11-16

## Related pages

- [Output Filtering](/glossary/output-filtering/)
- [Jailbreak](/glossary/jailbreak/)
- [Reverse Proxy](/glossary/reverse-proxy/)
- [Runtime Protection for Customer-Facing LLM Apps](/guides/runtime-protection-for-customer-facing-llm-apps/)
- [Runtime Security for RAG Pipelines](/guides/runtime-security-for-rag-pipelines/)
- [All glossary terms](/glossary/)
- [Inline Enforcement](/glossary/inline-enforcement/)
- [Guides](/guides/)
