After weeks of consideration and public backlash, the San Francisco Board of Supervisors came to a conclusion on the controversial subject of “killer robots” yesterday afternoon. They voted to temporarily halt the policy, which includes (with numerous amendments) explicit authorization for the San Francisco Police Department to use remote-controlled robots for deadly force.
The first vote of 8-3 last Tuesday was the first time that SFPD would’ve received such an approval.
At a rally on Monday morning in front of City Hall, District 5 Supervisor Dean Preston noted that the proposed policy was published on SFPD’s website on a Friday, three days before Monday’s hearing. This could violate Assembly Bill 481, which requires local law enforcement agencies to submit documentation of a proposed military equipment use policy on their website 30 days prior to any public hearing. The SFPD’s disclosure of “military equipment” included 17 total robots, an armored vehicle, and an LRAD device, among other tools.
It was the robots that received public attention, inspiring criticism from civil-liberty advocates who decried the policy as a disturbing milestone for police brutality in the Bay Area. The policy received national coverage from multiple media outlets in the lead-up to Tuesday’s meeting; even attracting derision from podcaster Joe Rogan in a widely shared Instagram post.
Here’s an overview of what occurred at last week’s and yesterday’s votes:
The Vote That Sparked Public Backlash
Last week’s vote came after six separate considerations of the policy by the Board of Supervisors’ Rules Committee, chaired by Aaron Peskin. Peskin had initially attempted to quash SFPD’s use of robots for deadly force outright, revising the department’s AB 481 disclosure to read that “Robots shall not be used as a Use of Force against any person.”
That suggestion was crossed out with a red line in a consequent response from the SFPD. At last week’s hearing, SFPD Assistant Chief David Lazar remained adamant that a robot would be used to kill a person in only the most extreme circumstances, invoking the 2017 Las Vegas shooting as an example of a situation that could require such a tool.
“These are not autonomous robots. There are trained officers who would utilize the robot. We would save lives,” Lazar told the Board. “We’ve had this capability and had this robot for 11 years … we use the robots every day, for de-escalation, for search warrants.”
Those assurances provided little relief for Supervisor Hillary Ronen, who described the pending authorization as a “Pandora’s Box” and expressed doubt on whether a robot could replace the judgment of a trained officer at the actual scene. She cited the SFPD’s history of problematic killings, as well as its failure to implement a lengthy list of reforms recommended by the federal government in 2016.
“So now there’s an idea that a robot can go in and do this work? There are so many reasons to argue against this … if you give someone a tool, the tool begs to be used,” Ronen said, adding that an officer’s remote operation of a robot means “you remove the emotional impact of taking an individual’s life.”
Preston was similarly incredulous about the Board’s consideration of an amended policy that would allow the use of deadly force by a police robot as a last resort. He also invoked the George Floyd protests of 2020, observing that keeping police accountable means pushing back against expansions of police power.
“I find it really shocking that two years after our nation collectively recognized that police were using unjustified deadly force, we’re having a conversation to allow SFPD to use robots to kill,” Preston said. “I think the City continues to write blank checks to the PD [police department] with very little oversight and very little justification … SFPD is 12 times more likely to use force against Black San Franciscans than white counterparts. Any tools you give to it, there will be disproportionate use against Black and Brown community members.”
The sharp criticism from Ronen and Preston along with more moderate disapproval from Supervisors Gordon Mar and Connie Chan, inspired pushback from the Board’s more openly pro-police representatives. A visibly annoyed Supervisor Rafael Mandelman took the mic to accuse civil-liberties advocates such as the ACLU of Northern California and the Lawyer’s Committee for Civil Rights, San Francisco of sensationalizing “killer robots.”
“Some of the rhetoric that has been used about this policy and generate debate, it leads me to feel that these organizations aren’t trustworthy for me to rely on. It feels like they’re just anti-police. That’s not where I am. That’s not where my constituents are,” Mandelman said. “I think there are problems when progressives, and progressive policy, look like it’s just anti-police. I think that’s bad for progressives, our city, and for Democrats.”
Supervisor Matt Dorsey took a more pragmatic angle, noting that he opposed “making my city a softer target” and suggesting that San Francisco should not “pioneer a restriction” on robot use-of-force that is different from that of other major cities. Supervisor Catherine Stefani, meanwhile, launched into an extended speech about how the militarization of police is necessary because of the “militarization of our society in the last 20 years,” decrying the spread of “assault weapons” and describing a list of mass shootings.
The initial language of the draft policy stated that “Robots will only be used as a deadly force option when risk of loss of life to members of the public or officers is imminent and outweighs any other force option available to SFPD.”
In last Tuesday’s amendments by the Board of Supervisors, the policy is read as such:
“Robots will only be used as a deadly force option when risk of loss of life to members of the public or officers is imminent, and officers cannot subdue the threat after using alternative force options or de-escalation tactics options, or conclude that they will not be able to subdue the threat after evaluating alternative force options or de-escalation tactics. Only the Chief of Police, Assistant Chief, or Deputy Chief of Special Operations may authorize the use of robot deadly force options.”
In the end, only Preston, Ronen, and Board President Shamann Walton voted against the amended policy.
Halting Killer Robots, For Now
Yesterday’s vote yielded a different result. After much deliberation on the language, 8 out of 11 Supervisors voted to amend the policy yet again, removing the authorization of the use of robots for deadly force. Supervisors Dorsey, Mandelman, and Stefani voted in dissent.
The Board also voted to send the original language on the use of robots for deadly force back to the Rules Committee, which might hold public hearings on the matter.
As of now, the Rules Committee will decide whether to approve those changes to the provision — by restricting the use of bomb-equipped robots with deadly force or banning the use of them entirely.
The Board is expected to take a final vote on the updated version of the policy next week.
“The people of San Francisco have spoken loud and clear: There is no place for killer police robots in our city,” Preston said in a press release after yesterday’s vote.
Prior to the second vote, Mar expressed regret about his first vote in a tweet, stating that he “grew increasingly uncomfortable with our vote and the precedent it sets for other cities without as strong a commitment to police accountability”.
Are Killer Robots the Solution?
It’s unclear whether an authorization of robots to use deadly force is a useful solution for the kinds of violence Dorsey and Stefani discussed. There is only one example of a “successful” robot killing in America: the 2016 killing of a Dallas sniper. Proponents claim that using robots could limit the danger to officers during standoffs and other violent events. But even beyond concerns of collateral damage and unnecessary deaths, critics of using potentially lethal robots say that the disturbing image of such a tool could further erode public trust of the police.
That’s a major reason the Oakland Police Department, after receiving feedback from the Oakland Police Commission and the public, decided last month to prohibit any use of a robot for deadly force. In San Francisco, however, the discussion only led to a decision to narrow, not restrict, how SFPD can deploy a deadly robot.
Yoel Haile, Director of the Criminal Justice Program at the ACLU of Northern California, made the following statement via email:
“No police department should be allowed to have killer robots. Police already shoot Black and Brown people with near impunity. Remote-controlled killing machines will not make San Francisco safer. They will make it easier for officers to mistakenly pull the trigger and impede people’s ability to seek justice when wrongful killings occur. This is a terrible idea, and it is what happens when our elected officials cave to the SFPD like they did when they gave them access to private surveillance cameras earlier this year. Police are encouraged to come back with more and more extreme asks.”
In some ways, nothing has changed: the SFPD had, for more than a decade, the ability to deploy a killer robot and didn’t. Looked at another way, these votes represent a real milestone in the city’s relationship to policing and technology, even if pro-police supporters say an armed robot is nothing to fear.
Savannah Kuang contributed to this story.
Photo by Savannah Kuang