San Francisco politicians to vote on policy endorsing lethal force for robots
Asimov would like a word
Next week, the San Francisco Board of Supervisors is expected to vote on a policy proposal that will allow the San Francisco Police Department (SFPD) to deploy robots authorized to kill people.
No such use is currently planned, according to SFPD, which characterized the policy proposal as an endorsement of the continued use of robotic systems it has acquired over the past eight years.
The proposal addresses California Assembly Bill 481, which requires law enforcement agencies to obtain approval from an appropriate legislative body for the use of military equipment.
To comply with the state law, the SFPD has developed a draft policy that supports its use of military-style kit and has been seeking the blessing of city supervisors. A vote on whether to adopt the proposal on first reading is scheduled [PDF] for November 29.
As reported by Mission Local, a San Francisco news outlet, Supervisor Aaron Peskin initially tried to exclude a lethal force option from the policy, but SFPD struck out the line he had added: "Robots shall not be used as a Use of Force against any person."
In its place, the policy now reads, "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."
Peskin did not immediately respond to a request for comment, but told Mission Local that the SFPD had argued there might be scenarios where lethal force was the only viable option.
- Dead or alive, you're camming with me, says RoboPup: Bomb squad hires Boston Dynamics Spot to snoop on suspects, packages
- So long, Top Gun... AI software waxes US F-16 pilot's tail 5-0 during virtual dogfight drills
- Boffins hunt and kill cockroaches with machine vision laser
- Hijacked, rampaging infrastructure will kill humans by 2025 – Gartner
Asked to explain why this policy change is need, a spokesperson for the San Francisco Police Department told The Register in an email:
SFPD has always had the ability to use lethal force when the risk of loss of life to members of the public or officers are imminent and outweigh any other force option available. Robots are often used to create distance between officers and subjects during some of the more dangerous and precarious situations our officers encounter. SFPD does not have any sort of specific plan in place as the unusually dangerous or spontaneous operations where SFPD's need to deliver deadly force via robot would be a rare and exceptional circumstance.
As reflected in the proposed use policy, SFPD officer's use of robots is guided by DGO 5.01, DGO 8.01, DGO 8.07, and DGO 8.08. The robots were acquired by the Department between 2010 and 2017 and have not faced a situation where the delivery of lethal force was necessary. The approval of this use policy allows the Department to continue using the tools in the current inventory in the manner we have since acquiring them. This use policy, required by AB 481, is the first time this Department has addressed the specific use of force via robots.
The robots identified in the draft policy include various remotely piloted robots designed for manipulating heavy objects, breaching walls, ordnance disposal, scouting, and surveillance. They include: Remotec modelsF5A, 6A, and RONS; QinetiQ Talon and Dragon Runner; IRobot FirstLook; and ReconRobotics Recon Scout ThrowBot.
None of these units are designed to carry firearms – a scenario considered and then abandoned by the Oakland Police Department – but at least some of them are capable of being used to kill.
In 2016, for example, Dallas police attached a pound of C4 explosives to a Northrop Grumman Remotec Androx Mark V A-1 robot and piloted the robot to a wall that a sniper – who had killed two police officers and wounded others – was using for concealment. Police then detonated the explosive, which destroyed the wall and killed the suspect, leaving the $151,000 robot with only minor damage to its arm.
Given a robot inventory of mainly bomb disposal bots and the fact that these are controlled by a remote human operator, the SFPD's policing looks unlikely to change much with a robot lethal force option.
That has yet to happen in San Francisco because these robots do not execute kill orders autonomously, without a human operator in the loop – the force policy language applies to "remotely controlled unmanned machine[s]". SFPD's robots are therefore not autonomous killer robots, but tele-operated tools. If they are used to inflict harm, such incidents should be evaluated for appropriateness like any use of force.
As for actual self-directed killer robots, if and when such devices become a practical and politically plausible option for US law enforcement, thyeir advent will create a more significant legal and technical transition. In the meantime, try not to die by automated car. ®