A Chinese Tech Worker at Zhihu Fired Allegedly via a Resignation Risk Prediction Algorithm

The firing of an employee at Zhihu, a large Q&A platform in China, was allegedly caused by the use of a behavioral perception algorithm which claimed to predict a worker’s resignation risk using their online footprints, such as browsing history and internal communication.

Risk classification

• Primary risk domain: 2 Privacy & Security
• Primary risk subdomain: 2.1 Compromise of privacy by obtaining, leaking or correctly inferring sensitive information

The Sangfor system compromised employee privacy by tracking and inferring non-public sensitive information, such as job search activities, resume uploads, and private chat logs, without explicit consent.

Additional risk subdomains

• 6.2 Increased inequality and decline in employment quality: The deployment of invasive AI surveillance tools in the workplace degrades the quality of employment and increases the power imbalance between employers and workers.

Causal factors

• Entity: Human
• Intent: Intentional
• Timing: Post-deployment

The privacy risks and workplace surveillance harms were caused by human decisions to develop, deploy, and utilize the monitoring system to track employees.

EU AI Act risk tier

• Risk tier: 2 High Risk

High Risk: The report describes an AI system used for employment and worker management, specifically for monitoring and evaluating employee behavior and predicting resignation risks.

AI system and alleged parties

• AI system: Sangfor Behavior Awareness System BA (Sangfor)
• AI purpose: Workforce Monitoring and Evaluation; Threat Detection
• Behaviour type: Assistant
• Alleged developer: Sangfor Technologies
• Alleged deployer: Zhihu
• Alleged harmed parties: Zhihu employees, Chinese tech workers

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Substantial. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Negligible, indirect Substantial
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Negligible, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Substantial, indirect Substantial
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Substantial, indirect Negligible
• Psychological: direct Negligible, indirect Minor
• Epistemic: direct Negligible, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Financial

Reported: The report describes the cost of the system (40,000 yuan for 3 years up to hundreds of thousands of yuan) and Sangfor's stock price drop, but does not report direct financial losses to victims other than the whistleblower's job loss.

Indirectly caused: The whistleblower suffered immediate financial loss from being laid off. Sangfor Technologies suffered a significant drop in market value, with its stock price falling over 55% from its peak.

Inferred additional harm: Many other employees likely suffered lost wages due to being fired or passed over for promotions based on the system's predictions.

Civil rights

Reported: The report explicitly describes violations of personal privacy rights and labor rights under China's Personal Information Protection Law.

Directly caused: The system directly violated the privacy rights of employees by tracking their non-public online footprints and communications without explicit consent.

Indirectly caused: Employers used the system to infringe on workers' freedom to seek alternative employment by preemptively laying them off.

Inferred additional harm: Systemic violations of privacy and labor rights likely occurred across thousands of other companies using this software without informing their employees.

Privacy

Reported: The report explicitly describes severe privacy violations caused by the surveillance system.

Directly caused: The system monitored and collected employees' private browsing histories, resume uploads, and chat logs on the company network without their informed consent.

Inferred additional harm: Millions of employees across Sangfor's 100,000 clients likely had their personal data and online behaviors monitored and analyzed without their knowledge or consent.

Psychological

Reported: The report explicitly describes psychological distress and anxiety among employees due to invasive workplace surveillance.

Indirectly caused: The whistleblower experienced distress from being abruptly laid off and reprimanded by their boss. Additionally, Chinese workers condemned the technology as a form of psychological abuse.

Inferred additional harm: Widespread anxiety, paranoia, and stress are highly likely among millions of employees working under constant, invisible surveillance, though specific numbers are unquantified.

People affected

• Occurrences reported: 1
• People reportedly harmed: 1
• People reportedly exposed: 200

Potential causes

Management

• Prioritizing Efficiency over Trust: Management values productivity and retention tracking over employee privacy.
• Inadequate Ethical Risk Review: Sangfor and clients failed to assess the ethical impact of surveillance.

Technology

• Predictive Resignation Algorithms: Algorithms analyze web traffic and chat keywords to predict employee turnover.
• Intranet Traffic Monitoring BA: System captures deep logs of all online activities on office computers.

Data Inputs

• Unconsented Employee Data: System collects personal job search logs and resume downloads without consent.
• Keyword Flagging in Chats: Internal communications are scraped for negative phrases like 'low wage'.

Human Factors

• Managerial Overreliance on AI: Managers rely on automated risk scores to make firing or layoff decisions.
• Employee Psychological Stress: Constant monitoring creates a climate of fear, distrust, and burnout.

Process and Methods

• Lack of Employee Notification: Companies fail to inform staff about the active behavioral monitoring tools.
• Flawed Risk-Rating Methods: Simplistic categorization of employees as 'high-risk' based on web visits.

Regulatory Environment

• Weak Intranet Privacy Laws: Lax enforcement of personal information laws regarding corporate intranets.
• Lack of Algorithmic Oversight: No regulatory review of behavioral prediction and employee ranking systems.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide consistent, detailed accounts of the Sangfor system's capabilities, pricing, patent history, and the public backlash. While Zhihu denied using the system, the existence and functionality of the software itself are well-documented and confirmed by Sangfor's own board secretary and sales representatives.
• Ambiguities identified: Whether Zhihu actually deployed the system or if the whistleblower's layoff was directly caused by this specific software remains officially denied by the company.
• Alternative interpretations: The system could be viewed purely as a network security and data leak prevention tool rather than an AI-driven employee surveillance system, as argued by Sangfor's board secretary.

The incident involves public controversy in China over Sangfor's AI-driven employee monitoring system designed to predict resignation risk. While raising significant privacy and labor rights concerns, its national security implications are minor and domestic, primarily managed through existing legal and regulatory frameworks.

• Overall national security impact: Minor
• Scope: Single nation
• Primary target: China
• Alleged perpetrator: Corporate employers

Threat characteristics

• Imminence: Long-term. The incident reflects ongoing corporate surveillance practices rather than an active national security crisis.
• Autonomy: Human-supervised. The AI system analyzes data and predicts resignation risk, but human managers ultimately make the decision to act or terminate employment.
• Novelty: Evolved capability. Represents an advanced application of existing employee monitoring and data analytics technologies to predict human behavior.

Impact by dimension

• Physical security: Negligible. No physical systems, kinetic threats, or critical infrastructure operations were affected or targeted in this incident.
• Information security: Negligible. No evidence of state-sponsored intelligence compromise, classified data theft, or information warfare campaigns.
• Sovereignty: Negligible. The incident involves domestic corporate surveillance and does not threaten state authority or core government functions.
• Economic security: Negligible. While the developer suffered a stock price decline, the incident does not pose a threat to national economic stability or strategic technological competitive advantage.
• Societal stability: Minor. The system represents invasive workplace surveillance affecting privacy and labor rights, sparking public backlash, but impacts are manageable within standard legal and regulatory frameworks.