Data broker startups are aggressively purchasing internal corporate files, emails, and professional work products to help train AI models. These transactions involve millions of dollars and a rapidly growing market, but not everyone is comfortable with these data acquisitions.
- Data brokers buy private corporate files and emails for AI training.
- Major startups claim to anonymize data before sale.
- Market predicted to grow from billions to over a trillion dollars annually.
What happened
Several emerging startups specializing in data brokerage are actively purchasing internal corporate workflows, documents, and emails to supply large AI labs and companies. Offers have ranged up to millions of dollars for access to extensive data collections. These companies acquire a variety of materials including professional documents and academic papers, which they claim help train AI models on realistic, complex work tasks rather than simulated data.
Notable players such as Handshake, Micro1, and Mercor have secured partnerships with many leading AI laboratories, fueling a substantial boom in this hidden industry. Mercor’s founders became billionaires within a short time after founding the firm in 2023, reflecting the rapid growth and scale of this data acquisition business focused on real-world workflow data.
Why it matters
As AI systems demand far more high-quality and diverse data, buying private corporate files has become as critical as investing in compute infrastructure. Industry executives estimate current spending on data acquisition at tens of billions annually, with forecasts projecting growth to over a trillion dollars in the next few years. This data market is thus becoming a core foundational layer for AI development.
However, these developments raise significant privacy and ethical concerns. Data brokers assert they anonymize or scrub personally identifying information before resale, but the sheer scale and intimate nature of corporate workflows being commodified prompts debate on the potential impacts to worker privacy, corporate confidentiality, and data security. The opacity of these data transactions also complicates public understanding and regulatory oversight.
What to watch next
Close scrutiny will be needed on how effectively data brokers can truly anonymize sensitive corporate information and whether legal frameworks adapt to this new class of data transactions. The growth trajectory of these companies and their partnerships with AI labs will shape how AI models evolve with access to increasingly realistic work data.
Regulators, corporate data owners, and AI developers should watch for emerging standards around consent, data rights, and privacy protections. The potential proliferation of data buyouts involving worker-generated content and corporate emails suggests a need for greater transparency and possibly new rules to balance innovation benefits against risks to individuals and organizations.