The MTA is using AI to track fare evasion on the subway

According to NBC News, the MTA is now using surveillance software that uses AI to spot people evading fares in the subway. The system analyzes videos from some of the system’s 10,000 cameras, which are stored temporarily on the MTA’s servers.

This new AI surveillance software was used in seven stations in May, a new MTA report says.

To be clear, the system isn’t tracking who evades fare nor does it alert the police. But according to NBC News, an MTA spokesperson declined to comment on whether it ever will.

“We’re using it essentially as a counting tool,” said Tim Minton, the MTA’s communications director. “The objective is to determine how many people are evading the fare and how are they doing it.”

The MTA says it lost $285 million to fare evasion on the subway in 2022—up to 13%, a ”sharp jump” from before the pandemic when it was just 3-6%. It also estimates that about 400,000 people jump, back-cock or duck under the turnstile or enter the emergency gate to evade the fare on an average weekday—enough people to fill Yankee Stadium eight times, it says.

Hiring human checkers to do a count of all evasion across the system would be too costly since there are 472 stations and about 1,000 fare collectors, the agency notes. This can only be done randomly by about 10 staffers each quarter. 

At the eight turnstiles using this new AI technology, the MTA found the following:

  • The evasion rate was about 16%, which is higher than the system-wide rate
  • More than half of the subway evasion was done by people simply walking through emergency exits
  • The largest spike in evasion usually occurred from 3 to 4pm

Seeing the usefulness of this information, the MTA says it’ll expand this new tech to about two dozen more stations, with more to follow. 

Source: Shaye Weaver, Editor, Time Out New York

Click here for the link to the full story The MTA is using AI to track fare evasion on the subway (


Enjoy this blog? Please spread the word :)

%d bloggers like this: