Flock Safety's AI Cop Search Tool Logs Abuse But Doesn't Stop It
Written by AI from the linked story. Not reviewed by an editor. How this works
Flock Safety's AI-powered surveillance tool allows law enforcement to search for individuals using natural language descriptions across a wide camera network, but code analysis reveals that policy violations are logged rather than actively blocked. This represents a critical gap between perceived safeguards and actual technical enforcement — a system that records misuse after the fact provides no real protection for individuals whose data is improperly queried. The tool's handling of sensitive attributes like race and religion raises significant civil liberties and bias concerns, as AI accuracy issues in these categories can lead to wrongful targeting. When powerful surveillance technology is deployed without hard technical controls, organizational policy alone is insufficient to prevent abuse.
Tactical Insight
Immediate actions
- Audit all AI-powered surveillance tools to verify that stated safeguards are technically enforced, not merely logged after the fact.
- Implement hard access controls (e.g., role-based query restrictions) that actively block prohibited search categories such as race or religion.
Policy & Governance improvements
- Require independent third-party code audits of any AI tool used in law enforcement before deployment.
- Establish a transparent public-facing use policy with mandatory review boards to oversee AI surveillance system queries.
- Define and enforce strict data retention limits and permissible use categories through binding legal agreements with vendors.
Detection & Accountability measures
- Deploy real-time alerting on anomalous or high-risk query patterns to a dedicated oversight team, not just passive logging.
- Require mandatory documentation and supervisor approval for any search involving sensitive demographic attributes.
- Conduct regular bias audits of AI model outputs to assess accuracy disparities across demographic groups.