Filestack Vice President of Product Sundip Patel and CEO Sameer Kamat Credit: Scott Ball / Rivard Report

San Antonio tech firm Filestack is taking a byte out of fake news.

The startup’s service helps about 100,000 app developers manage the flow of user-uploaded content, such as images, videos, and other files. That now includes the ability to seamlessly filter out doctored images.

Machine learning has recently been used in online dating platforms such as Bumble to block users from uploading photos that feature firearms.

Should a social media site such as Facebook or Twitter decide it wants to stamp out digitally manipulated images it could theoretically use Filestack to bar users from uploading forged photographs.

“Content is becoming so diverse that if businesses don’t figure out how to harness it, it’s going to be an extremely big challenge,” CEO Sameer Kamat said. “Images are just one thing.”

Filestack, which moved to San Antonio’s Houston Street tech district after it was bought by equity firm Scaleworks, has been evolving its role to include processing and curating content, Kamat said. Its application programming automatically checks for copyright infringement and explicit content.

Filestack provided this example of a manipulated image: The original image (left) has been altered into the images on the right, where more tanks have been added into the frame. Credit: Courtesy / Filestack

In a recent blog post, the tech startup detailed how its application programming interface, or API, can detect when the copy-move tool, one of the most common ways to digitally manipulate photos via software programs such as Adobe Photoshop, has been used. The API, which can be plugged into various software applications, can then remove counterfeit images to prevent spreading so-called “fake news” content.

Kamat said manually tagging, categorizing, and organizing content can be error-prone and expensive. Filestack’s goal is to make the processing of content online as close to 100 percent automated as possible.

“Let us do the heavy lifting with respect to content while you focus on your business,” he said. “If some business has to do this work themselves of uploading the file – transformation, curation, delivery – it’s going to take them months if not years to get it right, and they’ll spend millions of dollars doing it.”

Filestack’s API can integrate with software through a line of code, Kamat said. From there, it will return the result: Was the uploaded image manipulated? Is it inappropriate or “safe for work”? Is there a copyright infringement?

Filestack was founded in 2012 by MIT graduates. Scaleworks acquired the firm in late 2015, and the team now comprises 35 full-time employees.

The Filestack offices at Scaleworks. Credit: Scott Ball / Rivard Report

Most of its business comes from companies in education technology, print, and retail. Recently, an increasing number of drone companies have been employing the Filestack API, Kamat said, as drones are constantly uploading images, many of them large video files, from often remote areas with poor network connections.

The Filestack API can detect whether the network around the user is slowing down the upload. The API changes its algorithms, or sets of computer-programmed instructions, dynamically according to the congestion in the network to make uploading content faster, Kamat said.

About 2.5 quintillion bytes of data are created every day. As content on the internet mushrooms, Filestack wants to make the uploading files as seamless as possible.

“Content is changing,” Kamat said. “Content is almost doubling every 18 months or so. There’s a divergent curb of the growth of content and the network speeds, so the intelligence portion of our uploads becomes extremely important.”

JJ Velasquez

JJ Velasquez is the Rivard Report's technology reporter.

Read more