This is a research prototype. The data and analyses are preliminary and not yet validated — we'd welcome your .
Uber developed and deployed a secret program called 'Greyball' that utilized data analytics to identify and evade law enforcement and regulatory officials. By tagging these individuals, the system provided them with a fake version of the app showing 'ghost cars' to prevent them from successfully hailing rides during sting operations. This systematic deception was used globally to circumvent local regulations and avoid legal enforcement actions.
Uber developed a secret program "Greyball" which prevented known law enforcement officers in areas where its service violated regulations from receiving rides.
Domain classification, causal taxonomy, severity scores, and national security assessments were LLM-classified and may contain errors.
Using AI systems to gain a personal advantage over others such as through cheating, fraud, scams, blackmail or targeted manipulation of beliefs or behavior. Examples include AI-facilitated plagiarism for research or education, impersonating a trusted or fake individual for illegitimate financial benefit, or creating humiliating or sexual imagery.
Also classified as
The incident represents a major clash with regulatory oversight where Uber actively bypassed and undermined local transport governance frameworks.
Uber used the Greyball system to gain a commercial advantage over regulators and competitors by cheating, evading local laws, and performing targeted manipulation of the app's interface to deceive enforcement officers.
Human
Due to a decision or action made by humans
Intentional
Due to an expected outcome from pursuing a goal
Post-deployment
Occurring after the AI model has been trained and deployed
The risk was caused by the deliberate decision of Uber's management and legal team to deploy the Greyball tool to intentionally evade law enforcement after the app was deployed.
Added categories peak at 2: Minor (Epistemic Harm, Direct)
Uber utilized its secret 'Greyball' program to systematically identify, profile, and deceive local law enforcement and regulatory officials globally. While the incident represents a significant, coordinated effort to bypass government regulatory authority and compromise the privacy of targeted officials, its national security implications remain minor as it did not threaten state-level sovereignty, critical infrastructure, or national economic stability.
The Greyball system operated automatically to analyze user data (such as geofencing and credit card details) and make real-time decisions to tag users and serve fake app interfaces without requiring direct human intervention for each action.
Unacceptable Risk: The report describes an AI-driven system used for systematic deception and behavioral manipulation to evade law enforcement and bypass public safety regulations, posing significant dangers to society and the rule of law.
The primary goal of the Greyball system was to identify specific individuals (law enforcement and regulatory officials) using data analytics and geofencing, and model their behavior to serve them a fake version of the app.
The reports from The New York Times and The Guardian provide highly detailed, consistent, and well-sourced accounts of the Greyball program, including internal documents, statements from Uber, and reactions from local officials. The operational mechanics of the tool are clearly explained, leaving little ambiguity about its purpose and impact.
None. The evidence of intentional evasion of law enforcement is overwhelming and acknowledged by Uber's own statements defending the program.
Technology
Greyball Software Tool
Algorithmic tool created to identify and evade regulatory officials.
Ghost Car Simulation
Served fake version of app with virtual cars to mislead users.
Geofencing Algorithms
Monitored app usage near government offices to target officials.
Data Inputs
User Geolocation Data
Tracked location data to identify users near government offices.
Credit Card Information
Analyzed credit cards for links to police credit unions.
Device Fingerprinting Data
Identified cheap mobile devices commonly used in sting operations.
Human Factors
Driver Compliance with Cancellations
Drivers canceled rides on instruction to avoid sting operations.
Employee Ethical Concerns Ignored
Staff voiced concerns about the tool but were overridden.
Process and Methods
VTOS Target Identification Process
Systematic multi-step process used to flag and tag suspect users.
Greyball Playbook Distribution
Engineers distributed tactics playbook to international managers.
Social Media Profiling
Manual and automated online searches to confirm user identity.
Management
Legal Team Approval
Uber legal team approved the use of Greyball to evade law.
Aggressive Market Expansion
Prioritized rapid growth and market dominance over compliance.
Executive Awareness
Top executives were aware of and supported the evasion program.
Regulatory Environment
Bans on UberX Service
Local regulations banned low-cost services, prompting evasion.
Active Sting Operations
Law enforcement used undercover stings to impound Uber vehicles.
AI system involved
Identified by automated extraction; can differ from the alleged developers above.
