As AI laboratories compete to automate white-collar jobs, they have found an ideal source of training data: your Slack private messages.
Tech giants have begun collecting employee data for AI training. Moreover, intermediaries that originally only targeted the data of failed startups are now also paying to obtain data from companies that are still in operation.
"The current demand is simply astronomical," said Dan Novaes, the CEO of Mode.Inc, a startup that sells data to the developers of the AI model.
As companies such as OpenAI, Anthropic, and Meta create AI intelligents for businesses, ordinary office data has become more valuable than ever before. Gartner estimates that the annual expenditure on AI data is expected to soar from less than $1 billion last year to nearly $6.5 billion by 2027.
There is also a dark side to this craze. Employees rarely benefit from the sale of their Slack, PowerPoint, and Google Docs to the highest bidder, and in many cases, they are not even aware of it at all. It is suspected that hackers have sold the stolen codebase of a large company for millions of dollars. Meta's attempt to collect data from its own employees for AI model training has also triggered internal backlash.
Some CEO were caught up in it without any preparation. The CEO of Springshot, which operates an airline, Doug Kreuzkamp, learned during a Zoom meeting in August while sitting in his office that Google had purchased his data. Springshot has provided services for the now-defunct Spirit Airlines for a long time, helping to coordinate flights with gate staff and cleaners, among other tasks.
“I only found out in less than 48 hours – and still from the media – that the intellectual property we’ve spent 15 years building might be given away without a fight,” Kreuzkamp told Business Insider.
Emory University law professor Ifeoma Ajunwa stated that this rapidly expanding market is entering unknown territories, and the industry may soon come into direct conflict with employees and regulatory agencies. She added that due to the lack of federal-level privacy laws, companies face almost no legal obstacles when collecting employee data for AI training purposes.
"The issue is whether the consent obtained from employees when collecting this data actually extends to the sale of that data as well. I believe this falls into a grey area ethically," said Ajunwa.
Enterprise data has become the new oil.
This craze has given rise to a group of intermediaries who specifically target small businesses with abundant data but tight cash flow. After removing sensitive information such as social security numbers and home addresses, they resell this data to the AI giants.
Small business owners who hold a large number of internal records in their hands may not be aware of the considerable value of these seemingly ordinary documents. Lock says that a 48-hour quotation deadline has become quite common.
“These quotes look almost identical to those from early oil trades,” he said from his home in Dallas, “They would say, ‘I’ll offer you twice the market price; I need your reply by Friday.’”
Lock and his co-founders have also gone through the data authorization process themselves. In other companies, they sometimes couldn't help but laugh at the low quotes they received. Their company attempts to bridge this information gap by making it clearer for business owners to understand the value of data.
"There is still enough money in the market for everyone to get a share of it, but business owners should receive more profits," he said.
Requirement: Data from the company, regardless of whether it's alive or dead.
AI, the chief revenue officer of training startup Micro1, stated that data that can demonstrate how work is actually carried out is of great value, as it provides crucial context regarding how decisions are made and how products are built.
This information is helpful for training intelligent agents, which are those AI tools that can use computers autonomously. Companies such as Meta and SpaceXAI have already made this a focus. As AI companies build new models, real-world data will guide AI to handle specific tasks better.
Micro1 announced a plan earlier this year to pay medium-sized companies between $100,000 and $2 million in exchange for access to anonymized data. Almeida stated that AI developers are increasingly seeking "long-term" data that can fully represent work processes. If a company is unwilling to sell its Slack, customer relationship management, or email records, then the value of the remaining datasets will also decrease.
Data pricing depends on a variety of factors, among which scale is one of the key considerations. Mode.Inc CEO Novaes stated that the more white-collar employees a company has, the more valuable its dataset becomes.
Other factors are also important, such as exclusivity – if a competing laboratory has already obtained a certain code library, their demand for it will be lower. Areas like law and accounting, which are AI giants are most eager to automate, have particularly strong demand.
However, overall speaking, as long as there is sufficient data, any company – regardless of its status (alive or dead) or size – can sell its data.
"People need to wake up and seize this opportunity," said Novaes.
Grant Gooley decided to authorize the data of its own company to a AI training company. Its institution, MyPortal Marketing, has already sold various materials to Micro1, ranging from files from Figma and Canva to client onboarding lists, and even to Google advertisements. He stated that no customer-specific materials were sold.
“If AI is destined to be trained, then why not participate in it, and do so in a responsible manner?” said Gooley.
New concerns about security and privacy
While this market has created wealth for some people, it has also brought new cybersecurity and privacy risks to others.
At AI, they train startup Surge AI. Nick Heiner helps to improve AI intelligents by creating a digital environment similar to an office. He said that earlier this year, a group of people approached him offering the codebase of a large company for several million dollars.
“It quickly became apparent that this group of people did not have the right to sell this data,” Heiner told Business Insider, “They obtained an illegal copy of a code repository and then went around looking for data suppliers, asking, ‘Can this be sold for millions of dollars?’”
Concerns regarding security and consent do not come solely from hackers. The plan by Google to acquire the old data of Spirit Airlines for $10 million has encountered some obstacles, one of which is due to flight attendants' concerns about how their records might 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.
The spokesperson, Google, told Business Insider that the company has no intention of purchasing any personal information from Spirit Airlines; any such data will be excluded or undergo anonymization by an independent third party. Spirit stated that the company is working closely with flight attendants and other unions to address their concerns.
Another former freelance worker from a AI training startup, Mercor, has accused the company in a still-pending lawsuit of encouraging them to upload work materials without the employer's consent. Mercor denied these accusations in court.
The desire for work records is so strong that some tech giants have begun to track their own employees.
At Meta, when the company announced a new plan earlier this year to track employees' keyboard strokes and mouse movements for AI training, there was a backlash among the employees. They were concerned that Meta would capture their sensitive data, such as credit card numbers and immigration documents.
Although Meta Chief Technology Officer Andrew Bosworth promised that this data would be "under strict control," it turned out that this data could be viewed by the entire company during a major internal leak. Screenshots obtained by Business Insider showed that these so-called anonymized data contained employees' private conversations and performance records. Meta suspended the project not long after the leak and has not restarted it since then.
Business Insider Previous reports stated that Google considered tracking employees' AI outputs, but later abandoned this idea.
The internal rebound within Meta did not stop other companies from continuing to regard employee data as a shortcut to gaining a dominant position in AI, including the world's richest person.
SpaceX CEO Elon Musk told employees at a company-wide meeting in August that SpaceX would train its Grok AI using employee data, and encouraged employees to see themselves as the “parents” of this model.
“To some extent, it will be trained on you. It will inherit your thoughts, opinions, and beliefs, and I think that’s a good thing,” said Musk.
A vast market that is rarely recognized by employees
Although the company's data market is still in its early stages, AI training startups that profit from it are full of optimism.
Micro1 CEO Ali Ansari recently posted on X that granting data to laboratories could become one of the most common sources of revenue for global enterprises. He estimates that in the near future, about 10% of GDP in the United States may be used for AI model training due to corporate data and human professionals.
Last week, Micro1 announced a commitment to invest $1 billion in the company's data over the next 12 months.
Sociologists and former Uber policy executives, such as Alex Rosenblat, indicate that employee records are increasingly being used for AI training, which may lead to greater calls for new data protection and privacy regulations.
She said that this could also enable employees to gain new bargaining chips and negotiating power as the value of their daily digital footprints continues to rise.
"The problem is not that data collection didn't exist before; rather, the scale at which it can occur and the scale at which it becomes useful has changed," said Rosenblat.
"This is also an opportunity for people to say, 'Wait a moment, these values are slipping away from our hands.'"
The AI Data Grab
Slack
AI
More
Big Tech
Tech
Jump to
Main content
Search
Account
Jump to top of page
