US 12,386,924 B1
Deriving input restrictions for artificial intelligence agents
Rahul Parwani, Atlanta, GA (US); and Rohit Pradeep Shetty, Bangalore (IN)
Assigned to Airia LLC, Alpharetta, GA (US)
Filed by Airia LLC, Alpharetta, GA (US)
Filed on Nov. 13, 2024, as Appl. No. 18/946,880.
Application 18/946,880 is a continuation in part of application No. 18/885,556, filed on Sep. 13, 2024.
Claims priority of provisional application 63/658,434, filed on Jun. 10, 2024.
Int. Cl. G06F 21/12 (2013.01); G06F 16/3332 (2025.01); G06F 21/62 (2013.01)
CPC G06F 21/121 (2013.01) [G06F 16/3332 (2019.01); G06F 21/6245 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for controlling client-application-initiated execution of artificial intelligence (AI) pipelines, comprising:
receiving a request to generate a derived AI pipeline input restriction from a file that includes an example restricted input, the generation request being received from an administrative user of an AI pipeline management service, and the file being received through an upload to a first user interface (UI) provided by the AI pipeline management service;
causing a derived restricted input marker to be identified based on an evaluation of the example restricted input of the file, wherein the derived restricted input marker includes a content subject;
generating the derived AI pipeline input restriction by causing the AI pipeline management service to associate the derived restricted input marker with a management profile;
identifying a request from a client application to initiate execution of an AI pipeline, the request comprising a received input from the client application, wherein the received input originates from an end user through a second UI generated by the client application; and
enforcing the derived AI pipeline input restriction on the received input to the AI pipeline, wherein the enforcement includes at least one of blocking and modifying the received input prior to the received input reaching an AI model that is identified as part of the AI pipeline, and wherein the enforcement occurs in an instance in which the received input meets a threshold similarity to the content subject.