Exposing.ai

About

Exposing.ai is created by Adam Harvey and Jules LaPlace and is based on the earlier MegaPixels project (2017 - 2020). This new project, Exposing.ai, launched in January 2021 and is based on years of research about image training datasets used for face recognition and related biometric analysis technologies. After tracking down and analyzing hundreds of these datasets a pattern emerged: millions of images were being downloaded from Flickr.com where permissive content licenses are encouraged and biometric data is abundant. Telling the complex story of how yesterday's photographs became today's training data is part of the goal of this ongoing project.

During research fellowships with the Karlsruhe HfG Critical AI Group and Weizenbaum Institut in Berlin during 2020, Adam and Jules developed the technology that could allow users to check for their Flickr photos within these datasets. They then teamed up with Liz O'Sullivan and the team at the Surveillance Technology Oversight Project (STOP), and combined efforts to bring this project to the public. In January 2021 Adam, Jules, and the STOP team launched the new Exposing.ai website.

If you are Flickr.com user and have uploaded photos containing faces or other biometric information between 2004 and 2020, your photos may have been used to train, test, or enhance artificial intelligence surveillance technologies for use in academic, commercial, or defense related applications.

Exposing.ai provides a search engine to check if your Flickr photos were used in dozens of the most widely used and cited public face and biometric image datasets used for these purposes. Although Exposing.ai searches millions of records, it is not a fully comprehensive search for all image training datasets. Only datasets that were publicly accessible, include Flickr images, and included metadata are searchable on Exposing.ai. Countless more face recognition training datasets exist and are continuously being scraped from social media, news, and entertainment sites. Future versions of this project may expand to include more search options.

You can read more about the project on the FAQ page. For technical inquiries or information about further datasets to include on this site, contact Adam Harvey on Keybase at https://keybase.io/exposing_ai. A secure and anonymous connection can be established to share tips.

 Adam Harvey Concept, research, analysis,  design, and computer vision. https://ahprojects.com
Adam Harvey
Concept, research, analysis, design, and computer vision.
https://ahprojects.com
 Jules LaPlace Information architecture, web application, and citation geocoding. https://asdf.us
Jules LaPlace
Information architecture, web application, and citation geocoding.
https://asdf.us

Acknowledgments

The Exposing.ai project gratefully acknowledgment the guidance, feedback, and discussions with everyone who has been a part this project. In particular Liz O'Sullivan, Albert Fox Cahn, Matteo Pasquinelli, Nanna Bonde, Bianca Herlo, Daniel Kessler, and Marek Tuszynski (Tactical Tech).

Software Credits

Most all of this project was developed using open-source software.

Data analysis and web application development utilizes Python, Pandas, LibreOffice, Markdown, OpenCV, matplotlib, ReactJS, and many other great open-source software packages that were freely available. Project communication relied on Signal and Keybase. Face blurring uses the open-source VFRAME redaction software (disclaimer: also developed by Adam Harvey).

Cite Our Work

If you reference this research project or use any data from the Exposing.ai project, cite our original research as follows:

@online{Exposing.ai,
  author = {Harvey, Adam. LaPlace, Jules.},
  title = {Exposing.ai},
  year = 2021,
  url = {https://exposing.ai},
  urldate = {2021-01-01}
}