AI Dictionary › Prompting
Filtraggio dell'Output
Output filtering is the check applied to the text, image, or code generated by an AI model before it is returned to the user, aimed at detecting and blocking or correcting problematic content the model produced despite its safety instructions. It is a defense layer complementary to input-side filters: while those try to prevent dangerous requests, output filtering intercepts what the model actually generated, regardless of why it did so.
In practice, the generated text is analyzed by one or more secondary classifiers, or compared against lists of known patterns, before delivery. If a problem is detected, the system can fully block the response, replace it with a generic message, or ask the model itself to regenerate it under stricter constraints.
It is particularly relevant for use cases where the model generates unsupervised content in real time, such as public chatbots or image generators: an output filter reduces the risk that a single unexpected response goes viral on social media or exposes the company to reputational damage. It is also the tool used to enforce sector-specific rules, for example preventing a financial assistant from giving unauthorized investment advice.
The term became established with the large-scale spread of public chatbots starting in 2022-2023, when companies had to concretely address the risk of a model generating embarrassing or harmful content in production.
From our network
Magellano GPS: Fleet Tracking Made Simple
Real-time GPS tracking, remote engine lock, fuel and CO₂ reporting for your fleet.
Visit magellanogps.com →From the Agora Intelligence blog
📱 Download the Android app (beta) iOS coming soon
Say what you mean. Get what you need.
Grace Certified, the AI coach that trains and certifies your prompt engineering, by Agora Intelligence.