Knightscope Security Robots Get Pulled Across US Cities

The pitch for autonomous security robots was always tidier than the reality. In an investigation published this week, 404 Media, reporting in partnership with Proof News, tracked 21 US deployments of security bots from vendors including Knightscope and Oregon-based Daxbot and found that more than 13 have been quietly ended. Two of the AI experts in our Who's Who directory shared the reporting.

The failures rhyme. Dublin, Ohio pulled its "DubBot" from a downtown park after less than ten months of a planned two-year pilot, having produced no arrests or citations; city spokesperson Robyn Gray said it "did not fully meet our operational needs." San Antonio's international airport walked away from a proposed $21,000 one-year Knightscope lease after a one-month trial in which, per spokesperson Ana Flores, the robot "struggled with badge scanning, communication, and finding its way around." Salem, Oregon powered off three Daxbot units after a three-month pilot. New York City let Mayor Eric Adams's Times Square subway K5 program lapse in 2024, and the NYPD had already returned a Boston Dynamics robot dog back in 2021 after public outcry.

George Mason University robotics professor Missy Cummings told the outlet that "AI and robots for security have had a long problematic history," and that these systems perform "miserably" outside their training data. George Washington University law professor Andrew Ferguson described the machines as "a form of security theater, more optics than results."

Knightscope, founded in 2013, has accumulated $273 million in debt and has posted net losses every year since, according to the reporting. The company recently acquired security-guard firm Event Risk LLC, and CEO William Santana Li now sells a hybrid model, arguing that "the combination can be very powerful." Daxbot has pivoted too, contracting with at least nine cities since September 2025 for automated Americans with Disabilities Act sidewalk assessments, a narrower task where the off-training-data surprises that sank the patrol bots barely apply.

The piece does not total what municipalities collectively spent to reach this conclusion, nor does it break out how much of the $230 billion physical security market was ever plausibly addressable by autonomous units. For leaders watching AI-plus-hardware bets, the surviving deployments are the narrowly scoped ones. Sidewalk audits may quietly outlast the roboguard era they were supposed to define.