Boston City Council has voted to ban the use of facial recognition by the municipality, joining a growing list of administrations to outlaw the tech.

The decision comes amid a growing backlash against the software, which research shows consistently misidentifies people of color. An MIT study found that facial recognition algorithms designed by Microsoft, IBM, and Face++ made up to 35% more errors when detecting the gender of darker-skinned women. For light-skinned men, that error rate dropped was just 1%.

Councilor Ricardo Arroyo, who co-sponsored the law, said these biases made the tech unfit for use:

The law will prevent any city official from using the software or asking a third party to deploy it. However, they will still be able to follow up on tips from other law enforcement agencies found via facial recognition.

Boston joins the backlash

Boston’s vote makes it the second-largest city in the world to ban government facial recognition, after San Franciso, which last year became the first US city to outlaw the tech.

The city joins the Massachusetts municipalities of Springfield, Cambridge, Northampton, Brookline, and Somerville in banning the tech, following a sustained campaign from the ACLU. The Massachusetts branch of the civil liberties organization responded to the decision in a tweet :

At the state level, California, Oregon, and New Hampshire have all outlawed the use of facial recognition in police body cameras.

Even tech giants are starting to distance themselves from the software. In the wake of anti-racism protests sparked by the killing George Floyd, IBM, Amazon, and Microsoft all pledged to end or pause sales facial restriction . However, Amazon and Microsoft have only suspended sales to law enforcement, and both aim to resume them once federal regulations are introduced.

The Boston vote came the same day as a report that a black man in Michigan had been wrongfully arrested due to a faulty facial recognition match — the first known case of its kind.

Tomasz David
Tomasz David

Leave a Comment