Mystery Synthetic Text Detector 3000
U.S. flag

An official website of the United States government

Official websites use .gov
A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS
A lock ( ) or https:// means you’ve safely connected to the .gov website. Share sensitive information only on official, secure websites.

Focus Area:

Influence and Information Capabilities

Project Cost:

$745,000

Project Duration:

10 months

Performer:

Johns Hopkins University Applied Physics Lab

Deliverable:

This project will result in the development of a prototype, user-friendly, software platform to reliably detect the presence of synthetically generated text that is designed to erode public trust in reliable sources or for disinformation campaigns.

End User:

SOCOM J39, Department of State

Description:

Synthetic text generation capabilities currently outpace synthetic text detection capabilities, with commercially available detectors being imperfect, deceivable, and useful only to experts. A more robust system needs to be developed for use by a wide range of non-technical experts: analysts, open source intelligence (OSINT) collectors, government public affairs officers, Foreign Service Officers, and others within the interagency who work to combat disinformation but are not technical specialists. The government requires development of a reliable synthetic text detection application, with flexibility for upgrades, which can support operational experimentation while also being open to modification based on user feedback.

Operational Impact:

If successful, this project would represent a significant enhancement over current human (manual) analysis by increasing the speed and accuracy of detection.

Transition:

Base contract will provide end users with an updateable app or plug-in tool that will significantly enhance their analysts’ ability to identify and counter adversarial use of synthetic text during influence messaging operations, OT&E updates and feedback sessions, end user trainings and training manual.