Kong has launched a new version of the AI Gateway with improved AI governance functions.
One of the biggest challenges when working with LLMs is their tendency to ‘hallucinate’ or generate incorrect information. Kong tackles this problem with an AI RAG Injector. This feature automatically connects to a vector database to add relevant data to a prompt before sending it to the model.
By transferring responsibility for RAG pipelines to the AI Gateway, the vector database is placed behind an extra security layer, improving security and regulatory compliance. Developers benefit from a low-code/no-code approach, which means they spend less time manually building these pipelines.
Improved personal data protection
The new PII sanitization function protects personal data, passwords, codes and more than 20 categories of PII in 12 different languages. Platform owners can enforce this sanitization (the process of identifying, masking or removing PII) at the platform level, so developers do not have to implement it separately for each application.
A significant advantage is that, unlike other sanitization products that simply remove sensitive data, Kong AI Gateway 3.10 can optionally reinsert the original data into the response before it reaches the end user. This ensures a seamless and personalized user experience.