AI Companies Are Now Buying Your Private Work Correspondence

The pace of artificial intelligence development is so rapid that tech giants have already exhausted public internet resources. Consequently, labs like OpenAI, Anthropic, and others have turned their attention to a more intimate source: private company documentation. A network of intermediary startups has emerged, dedicated to acquiring and processing internal business communications, Slack messages, and emails to train AI models.

The Data Gold Rush

Mercor is considered a leader in this field, with its founders achieving billionaire status by the age of 22. The company actively aggregates databases and supplies them to major AI labs. Similar operations are being conducted by Micro1 and Handshake. The latter, for instance, purchases student assignments and various professional documents to help AI better understand how real-world businesses operate. Some experts describe this massive data collection process as a trillion-dollar market that will become an integral part of AI infrastructure in the coming years.

Privacy Risks and Ethical Questions

One of the most pressing issues in this data collection process is the protection of personal information. Companies claim they anonymise data and fully remove personally identifiable information (PII). However, experts remain sceptical. Emails and internal chats contain unstructured information where employees often discuss personal matters, health conditions, or family issues. Fully "cleaning" such data is practically impossible.

A clear example of this problem is the Spirit Airlines bankruptcy case, where an attempt was made to sell the company's digital assets, including millions of internal messages. This sparked fierce protests from flight attendant unions, who find the commercialisation of employee data for AI training unacceptable. The unions argue that this data is being valued solely to train AI models to replace human workers.

What This Means for the Future

Although startup representatives claim these efforts serve the "greater good," the reality is that employees' work is being used to build technologies that may eventually replace them. It is likely that in the near future, selling workflow data will become as common a revenue stream for companies as advertising is today.

For the Georgian market, these developments are not yet highly tangible, but the global trend is clear: data protection standards are failing to keep pace with technological ambitions. In the future, as Georgian companies become more actively involved in the international digital ecosystem, questions regarding the ownership of work correspondence will become critical for local businesses as well.