in

Inside America’s School Internet Censorship Machine


APS, which installed Blocksi in May, stopped using the filter on most of its devices in August due to its restrictiveness, Harris says, and returned to the GoGuardian filter it used before the switch. Our investigation raises questions about the appropriateness and implementation of GoGuardian’s filter as well.

In May, before the district switched to Blocksi, the GoGuardian filter blocked an 8th grader from searching for “suicide prevention.” It prevented a 3rd grader from searching the word “latina” and a 6th grader from searching “black man.” When an 11th grader Googled “Obergefell v. Hodges ruling,” instead of a list of websites with information about the landmark United States Supreme Court case that legalized same-sex marriage, the student saw a grey screen with APS’s logo and the message: “Restricted. This website has been blocked by your administrator.”

It is difficult to determine who exactly is responsible for a given content restriction. While APS administrators set the network policy for the entire district, individual teachers can also choose what to filter with GoGuardian—including whether to turn off the internet entirely for a particular student or class during a lesson, according to Harris. Outside of school hours, parents can also use the Blocksi and GoGuardian parent apps that APS provides to set their own restrictions on their kids’ school-issued devices.

Blocksi did not respond to multiple requests for comment or answer detailed questions about censorship of APS web activity.

Jeff Gordon, director of public relations for GoGuardian, tells WIRED, “GoGuardian regularly evaluates our website categorization to ensure, to the best of our ability, that legitimate educational sites are accessible to students by default.” He said more than 7,600 school districts use the company’s web filter and referred all questions about whether the blocked activity in Albuquerque was appropriately censored to the district.

Sithara Subramanian, an 11th grader at La Cueva High School, says she began to run into her school’s GoGuardian filter on a regular basis around the time remote learning ended. “It got kind of intense when we went back to school, like educational websites were being blocked,” Subramanian says. The censorship has been particularly frustrating for her biology and anatomy studies. “It felt like they were trying to restrict our education rather than enhance it.”

“My son says the filters make the internet useless,” Sarah Hooten, the mother of Henry, a 13-year-old former APS student, tells WIRED. Henry says that he couldn’t use YouTube to look up information for a report he was assigned about rainforests. “I know it’s partly to do with blocking kids from doing what they aren’t supposed to be doing,” Henry says. “But it’s also just the school not understanding what they are blocking.”

What Went Wrong

The scale of censorship we found in Albuquerque’s schools shows how web filters can twist seemingly simple decisions to block unwanted online content into policies that render the internet near-impossible to use.

In one instance, an APS staff member was unable to view The New York Times’ Pulitzer Prize-winning “1619 Project,” a historical exploration of slavery and its consequences in the United States, because of an apparently misguided keyword block in the district’s Blocksi filter. The district’s web-filter blocked websites containing the keyword “avery.” This blocked hundreds of attempts to access the website of a printing company, Avery.com, although APS officials could not explain why “avery” was keyword-blocked. But because the URL for the 1619 Project includes the word “slavery,” it was also blocked. So was a Stanford University lecture about slavery, a Wikipedia map of slavery in the United States, and several articles about a controversial Florida curriculum about slavery.

OpenAI CLIP: ConnectingText and Images (Paper Explained)

How Getir reduced model training durations by 90% with Amazon SageMaker and AWS Batch

How Getir reduced model training durations by 90% with Amazon SageMaker and AWS Batch