Chinese tech workers are starting to train their AI doubles–and pushing back

2026-04-20 · Source: MIT Technology Review · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Human Resources & Workforce Development, Operations & Process Management · Depth: Novice, medium

Summary

A viral GitHub project named "Colleague Skill" has sparked significant debate among Chinese tech workers regarding AI-driven job automation and worker dignity. Created by Tianyi Zhou of the Shanghai Artificial Intelligence Laboratory, the project, initially a spoof, allows users to "distill" a coworker's skills and personality from chat histories and files from apps like Lark and DingTalk, generating manuals for AI agent replication. This trend coincides with companies encouraging employees to document workflows for automation using tools like OpenClaw and Claude Code. While some find humor in the concept, others, like Shanghai tech worker Amber Li, describe the experience of replicating a former colleague as "uncanny and uncomfortable." Emory University's Hancheng Cao notes that companies gain internal experience and richer data on employee know-how, aiding in standardizing work processes.

Key takeaway

For CTOs and VPs of Engineering evaluating AI agent adoption, recognize that while these tools offer data on workflows and potential standardization, they also introduce significant employee alienation and legal ambiguities regarding intellectual property. Your teams should prioritize understanding the human element and ethical implications of self-automation, potentially exploring "anti-distillation" countermeasures to protect worker dignity and ensure AI integration supports, rather than diminishes, human value.

Key insights

AI tools are prompting tech workers to automate their roles, raising concerns about job security and worker identity.

Principles

• AI agent utility remains limited in complex business contexts.
• Documenting workflows provides firms with valuable operational data.

Method

The "Colleague Skill" project imports chat history and files from workplace apps to generate reusable manuals detailing a coworker's duties and quirks for AI agent replication.

In practice

• Use "anti-distillation" tools to obscure workflow documentation.
• Experiment with AI agents to understand their current limitations.

Topics

• Colleague Skill Project
• AI Agent Automation
• Worker Alienation
• Workflow Documentation
• Anti-Distillation Skill

Best for: CTO, VP of Engineering/Data, Director of AI/ML, Software Engineer, AI Product Manager, AI Ethicist

Related on AIssential

• Anthropic slams Chinese AI firms for harvesting data from its Claude chatbot — AI frontier models accessed via APIs present a new attack surface for industrialized data distillation.
• AI usage among white-collar workers and students in Hong Kong: survey findings — Hong Kong shows high AI adoption among workers, but a significant executive leadership gap and workflow integration challenge persist.
• Amazon Employees Are Faking Their AI Usage — Internal pressures can lead to inflated reporting of AI tool adoption and usage within large organizations.
• Meta just fired 7,800 employees and used their daily work to train AI — Meta trained AI on its employees' work before layoffs, indicating a strategy to cut costs and automate human roles.
• US accuses China of “industrial-scale” AI theft. China says it’s “slander.” — The US government is preparing to counter alleged industrial-scale AI intellectual property theft by Chinese entities using distillation attacks.
• The AI paradox: More automation, more humans, more work | Dan Shipper — AI paradox: automation increases human work, shifting tasks to agents and integrated work surfaces.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT Technology Review.