As AI labs compete to automate white-collar work, they’ve found an ideal source of training data: your Slack private messages.
Tech giants have begun collecting employee data for AI training. Meanwhile, data intermediaries that previously focused only on defunct startups are now paying for data from companies still in operation.
"The current demand is astronomical," said Dan Novaes, CEO of Mode.Inc, a startup that sells data to AI model developers.
As companies like OpenAI, Anthropic, and Meta build AI agents for enterprises, ordinary workplace data has become more valuable than ever. Gartner estimates that annual spending on AI data is expected to surge from less than $1 billion last year to nearly $6.5 billion by 2027.
This surge also has a dark side. Employees rarely benefit when their Slack channels, PowerPoint presentations, and Google Docs are sold to the highest bidder—and often don’t even know it’s happening. Suspected hackers have attempted to sell stolen code repositories from a major company for millions of dollars. Meta’s efforts to collect employee data for AI model training have also sparked internal backlash.
Some CEOs were caught off guard. Doug Kreuzkamp, CEO of Springshot, which operates airline services, learned during a Zoom meeting in August that Google had purchased his data. Springshot had long served the now-bankrupt Spirit Airlines, helping coordinate tasks between flight crews, gate agents, and cleaners.
“I found out—less than 48 hours ago, and from the media—that the intellectual property we spent 15 years building might be handed over,” Kreuzkamp told Business Insider.
Emory University law professor Ifeoma Ajunwa said that this rapidly expanding market is entering uncharted territory, and the industry may soon face direct clashes with employees and regulators. She added that, due to the lack of federal privacy laws, companies face few legal barriers when collecting employee data for AI training.
“The issue is whether the consent employees gave when their data was collected truly extends to the sale of that data. I think this sits in an ethical gray area,” says Ajunwa.
Corporate data has become the new oil.
This surge has given rise to intermediaries who specifically seek out small businesses with rich data but tight cash flow. After removing sensitive information such as Social Security numbers and home addresses, they resell this data to AI giants.
Small business owners holding large volumes of internal records may not realize that these seemingly ordinary documents are highly valuable. Lock says that a 48-hour quote deadline has become common.
“These quotes look almost identical to those from early oil trades,” he said in his home in Dallas. “They would say, ‘I’ll give you twice the market price—I need your response by Friday.’”
Lock and his co-founders themselves went through the data authorization process and sometimes laughed at the lowball offers they received from other companies. Their company aims to bridge this information gap by helping business owners better understand the value of their data.
“There’s still enough money in the market for everyone to get a share, but business owners should receive a larger portion of the profits,” he said.
Requirement: Data from companies, alive or dead
William Almeida, Chief Revenue Officer at AI training startup Micro1, said that data showing how work actually gets done is extremely valuable because it provides critical context on how decisions are made and products are built.
This information helps train agents—AI tools capable of autonomously using computers—that companies like Meta and SpaceX AI have prioritized. As AI companies develop new models, real-world data guides AI to perform specific tasks more effectively.
Earlier this year, Micro1 announced a plan to pay medium-sized companies between $100,000 and $2 million in exchange for access to anonymized data. Almeida said AI developers are increasingly seeking “long-cycle” data that fully captures workflows. If a company is unwilling to sell its Slack, customer relationship management, or email records, the value of other datasets also decreases.
Data pricing depends on multiple factors, with scale being one of the key factors. According to Novaes, CEO of Mode.Inc, the more white-collar employees a company has, the more valuable its data set becomes.
Other factors are also important, such as exclusivity—if a competing lab has already acquired a particular codebase, demand for it will be lower. Demand is especially strong in areas that AI giants most want to automate, such as legal and accounting.
However, overall, as long as there is sufficient data, any company—alive or dead, large or small—can sell its own data.
"People need to wake up and seize this opportunity," said Novaes.
Grant Gooley decided to license his company’s data to an AI training company. His organization, MyPortal Marketing, has already sold a variety of materials to Micro1, including files from Figma and Canva, customer onboarding checklists, and Google Ads data. He said that no client-specific materials were sold.
“If AI is destined to be trained, why not participate—and do so responsibly?” said Gooley.
New Security and Privacy Concerns
While this market has created wealth for some, it has also introduced new cybersecurity and privacy risks for others.
At the AI training startup Surge AI, Nick Heiner helps improve AI agents by building digital environments that resemble offices. He says that earlier this year, a group approached him offering access to a large company’s codebase for millions of dollars.
It quickly became clear that this group had no right to sell this data,” Heiner told Business Insider. “They illegally obtained a copy of the codebase and went around to data suppliers saying, ‘Can you sell this for millions of dollars?’”
Concerns about security and consent do not come solely from hackers. Google’s plan to acquire Spirit Airlines’ old data for $10 million faced partial obstruction, partly due to flight attendants worrying about how their records would be used. Last week, Senator Elizabeth Warren and more than 100 members of Congress jointly wrote to Google and Spirit, urging both companies to protect the privacy of thousands of former Spirit employees.
A Google spokesperson told Business Insider that the company has no intention of purchasing any personal information from Spirit Airlines, and any such data would be excluded or de-identified by an independent third party. Spirit representatives said the company is working closely with flight attendants and other unions to address their concerns.
A former gig worker at Mercor, another AI training startup, has alleged in an ongoing lawsuit that the company encouraged them to upload work materials without their employer’s consent. Mercor has denied these allegations in court.
The desire for work records is so strong that some tech giants have begun tracking their own employees.
At Meta, employees revolted when the company announced earlier this year a new initiative to track keystrokes and mouse movements for AI training. Employees feared Meta would collect their sensitive data, such as credit card numbers and immigration documents.
Although Meta’s Chief Technology Officer, Andrew Bosworth, promised that this data would be “strictly controlled,” it was revealed that the data was accessible to the entire company during a major internal leak. Screenshots obtained by Business Insider show that these so-called anonymized data included private employee conversations and performance records. Meta paused the project shortly after the leak and has not restarted it since.
Business Insider previously reported that Google had considered tracking employees' AI outputs but later abandoned the idea.
The internal rebound at Meta has not stopped other companies from continuing to view employee data as a shortcut to AI dominance, including the world's richest person.
At an all-hands meeting in August, SpaceX CEO Elon Musk told employees that SpaceX would use employee data to train its Grok AI and encouraged staff to view themselves as the "parents" of this model.
“To some extent, it will be trained on you. It will inherit your thoughts, perspectives, and beliefs, and I think that’s a good thing,” Musk said.
A vast market that few employees realize exists
Although the company data market is still in its early stages, AI training startups that profit from it remain optimistic.
Micro1 CEO Ali Ansari recently posted on X that licensing data to laboratories could become one of the most common revenue sources for global enterprises. He estimates that, before long, approximately 10% of U.S. GDP may stem from corporate data and human expertise used to train AI models.
Last week, Micro1 announced a commitment to invest $1 billion in company data over the next 12 months.
Sociologist and former Uber policy executive Alex Rosenblat said that the increasing use of employee records for AI training may intensify calls for new data protection and privacy regulations.
She said this could also give employees new leverage and bargaining power as the value of their daily digital footprints continues to rise.
“The issue isn’t that data collection didn’t exist before; it’s that the scale at which it can occur—and the scale at which it becomes useful—has changed,” Rosenblat says.
This is also an opportunity for people to say, “Wait, these values are slipping through our hands.”
The AI Data Grab
Slack
AI
More
Big Tech
Tech
Jump to
Main content
Search
Account
Jump to top of page