Block unsafe AI inputs and unapproved AI use

Stop prompt injection, data leaks and unapproved AI tools before they reach a model, and set the rules for who may use which AI. 44 products across 34 companies in the Ludbee catalogue carry this task — spanning every layer of the stack that exists here, not one of its own: 22 Application, 8 Infrastructure, 1 Foundation Models, 3 Hardware.

Filter the whole directory by this job to sort and compare these products.

Where these run

SaaS
33 of 44 products that record it
Self-hosted
12 of 44 products that record it
Private cloud
6 of 44 products that record it
Hybrid
5 of 44 products that record it
Air-gapped
2 of 44 products that record it
On-device
1 of 44 products that record it

One product can run more than one way — the counts above are of records, not of products.

What this page does not claim

This list carries no score and no recommendation, and one order only: inside each layer, a company whose product for this task is marked AI feature — sold for another job, with AI as one function its own page names — comes after the rest, A–Z within each. The standing rule that applies to the whole directory applies here too: no payment changes a figure, a ranking or the order of any list, this one included. Nor can it yet say what governs any of these products once deployed — what data they can reach, what they are certified against, whether a vendor trains on what you send it. That field is defined in the ontology and is populated on none of the 4,528 products in the catalogue yet, so a page that leaned on it would be promising an answer the data cannot back. It will say so here the day that changes.

The companies

Application (22)

Infrastructure (8)

Foundation Models (1)

Hardware (3)