Sound Transit wants to Install AI Cameras

Link light rail stations and garages would use new surveillance equipment


By George Howland Jr.

In the first quarter of the next year, Sound Transit security staff will ask the Sound Transit Board of Directors to take the first step in installing artificial intelligence (AI) cameras throughout its Link light rail system of stations and garages. The Sound Transit Board of Directors will be asked to approve a Request for Proposals (RFP) from private vendors for the initial $979,200 purchase of the AI technology. There are no public meetings or public outreach planned in connection with this decision, according to John Gallagher, Sound Transit spokesperson. Branden Porter, Sound Transit’s System Security Manager, states the objective of the AI cameras is to save over $1 million per year in security costs. Sound Transit’s Gallagher stresses, “We’re not using these [cameras] in any way for facial recognition.”

Critics are not mollified by the promise. Shankar Narayan, Technology and Liberty Project Director for the American Civil Liberties Union of Washington (ACLU-WA), says, the use of AI cameras would force “people to choose between privacy and transit.” This controversy comes on the heels of Sound Transit, last month, releasing data that shows its fare enforcement results in disproportionate citations for African Americans (first reported by Erica C. Barnett and then The Seattle Times).

Voters in a tri-county region (roughly approximating King, Snohomish and Pierce counties) have approved an ambitious expansion plan of Sound Transit’s light-rail, express-bus and heavy-rail transit services. Sound Transit’s current light rail is 22 miles long and runs from the University of Washington’s Husky Stadium to Angle Lake, just south of Sea-Tac Airport. By 2041, Sound Transit plans to expand light raid north to Everett, south to the Tacoma Dome and east to Issaquah. There are also two planned expansions in the city of Seattle itself: one to West Seattle and the other to Ballard. Recently, statewide voter approval of Initiative 976 put Sound Transit’s funding into doubt. Currently, I-976 is facing a court challenge.

According to Sound Transit’s Porter, current light rail expansion plans will result in 12 new stations by 2030. Under Sound Transit’s current system of staffing each facility would be monitored by security personnel 24 hours a day, 7 days a week at a cost of $28 million over ten years—2012-2030. By switching to the AI cameras, Sound Transit would only spend $15 million over the same time period. The AI cameras would be monitored by security guards who are “looking for patterns,” explains Sound Transit’s Gallagher. The cameras will alert the guards if someone is not moving or if people appear to be fighting, says Sound Transit’s Gallagher. He adds, “This is commonly used technology in security including transit systems.” Transit systems in Miami and San Francisco are already using AI surveillance, states Gallagher.

ACLU-WA concerns: overcollection of data, biases against people of color and no public process

ACLU-WA’s Narayan says, “We are concerned about surveillance technology overcollecting data, putting marginalized and vulnerable c ommunities in fear. You can’t find an example of surveillance technology that hasn’t been used against marginalized and vulnerable communities.” Narayan says it doesn’t matter that Sound Transit pledges not to use the facial recognition capability of the technology. “You can apply facial recognition after the fact. The federal government can do that now with its own tools,” he says. “If you have data being collected on this massive scale, there is going to be mission creep. It’s not a question of if, it’s a question of when.”

Both the Federal Bureau of Investigation (FBI) and the Immigration and Customs Enforcement (ICE) have accessed the data collected by state agencies to use facial recognition technology to investigate low-level crimes, according to The Washington Post. “Since 2011, the FBI has logged more than 390,000 facial-recognition searches of federal and local databases, including state DMV [Department of Motor Vehicles] databases…” The Post reported in July.

Narayan also argues that AI systems have built-in biases against people of color. “There is no perfect way to build a system like this. There are biases in the data set. Motion detection systems are less accurate for people of color. These systems put people into buckets. They are going to misclassify people just as a result of the way they are designed. It may well end up with consequences for people.” Given Sound Transit’s history of disproportionate fare enforcement, the agency should proceed with a great deal of caution, says Narayan. “This is an agency that already has a history of targeting people of color.”

In addition, Narayan is appalled that there is no public process planned before the decision to employ AI cameras. “We need to have a democratic debate about it,” he says. “The public hasn’t had an opportunity to weigh in. Where there are automatic detection systems, there should be transparency.”

I called King County Executive Dow Constantine, who is a member of the Sound Transit board, for comment. His office referred me to Redmond Mayor John Marchione, the current chair of the Sound Transit board. Marchione’s office says he is unavailable for comment.

Narayan says, “There is no excuse for policymakers not to take this seriously. We see a future where it will be impossible to live without being surveilled. It turns the presumption of innocence on its head.”

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s