AI Labs Are Aggressively Purchasing Private Corporate Workflow Data
Companies are buying emails and internal documents to train models, sparking significant privacy concerns.
1The Rise of Data Brokers
AI labs are aggressively acquiring private work records, including emails, Slack messages, and internal documents, to improve model performance. Startups like Mercor and Micro1 are leading this gold rush, paying millions to secure proprietary datasets. This industry boom aims to capture the messy, authentic nature of human professional workflows for training.
2Exhaustion of Public Data
The primary driver for this trend is the depletion of high-quality public internet data. AI companies are now shifting their focus toward private corporate information, which they believe is essential for building more capable models. Industry experts estimate that annual spending on private data acquisition could reach one trillion dollars soon.
3The Mercor and Micro1 Model
Mercor, founded in 2023, has become a dominant player by brokering deals between companies and major AI labs. Similarly, Micro1 is actively acquiring large datasets, even bidding on the digital assets of bankrupt entities like Spirit Airlines. These firms claim to strip personal identifying information before selling the data to developers.
4Privacy and Ethical Risks
Despite claims of anonymization, privacy advocates warn that unstructured data like emails often contain sensitive personal details that are difficult to remove. The Electronic Privacy Information Center has challenged the sale of employee data, arguing that anonymization cannot fully mitigate the risks of leaks or the exposure of private family matters.
5The Spirit Airlines Conflict
A recent dispute over bankrupt Spirit Airlines' digital assets highlighted the legal gray areas of this industry. The sale included millions of internal messages and files, triggering a fierce bidding war between Google and Micro1. Flight attendant unions have formally objected, citing concerns over the unauthorized commodification of their private information.
6Worker Consent and Automation
A troubling aspect of this data trade is that it often occurs without explicit worker consent. Employees are effectively contributing to the development of AI models designed to automate or replace their own roles. Critics argue that this practice turns human labor into a commodity used to build tools for displacement.
7New Revenue Streams Emerging
Industry leaders suggest that selling workflow data will soon become a standard revenue line item for corporations, potentially surpassing advertising income. While some executives view this as a necessary step toward future abundance, others fear it creates a permanent, involuntary pipeline of human knowledge into the maw of AI systems.
8Lack of Industry Standards
Venture capitalists and legal experts note that the AI industry is currently operating without established norms for data acquisition. This rapid, largely unregulated expansion is intentional, allowing companies to move faster than regulators can respond. The current landscape resembles a Wild West, where the value of data outweighs ethical considerations.
9Implications for Future Work
The commodification of human workflow signals a fundamental shift in how AI training data is sourced. As labs prioritize real-world operational data over synthetic alternatives, the tension between corporate profit and individual privacy will intensify. The long-term impact on the workforce remains a critical, unresolved issue for society and labor unions.
Original article: Wired — AI Labs Are Aggressively Purchasing Private Corporate Workflow Data