Lucrative internal data: AI firms buy internal data from liquidated companies
A veritable gold rush has developed in the USA around the liquidation of failed companies: AI firms are buying their internal communication data.
(Image: siro46/Shutterstock.com)
US companies that help with company liquidations can increasingly sell their internal communication to AI companies for high prices, who want to train their models with it. This is reported by the US magazine Forbes, citing a company that has already received more than a million US dollars for almost 100 such data sets. Typically, between 10,000 and 100,000 US dollars are paid for Slack archives, emails, source code, and more, according to a company called SimpleClosure. Particularly large sums are flowing for data sets from liquidated companies in the healthcare or financial sectors. All personal data would be removed from the data beforehand.
Authentic data of enormous interest
Forbes recalls that human-generated content is worth its weight in gold for AI training. However, content accessible on the internet – such as on Wikipedia or Reddit – has already been completely scraped since 2024. What employees write internally in the context of their work, on the other hand, is not only suitable for AI, which is intended to be used precisely there. AI companies also cannot access such content, as they cannot simply retrieve it. Company liquidations are therefore of particular interest, and a business field has now developed from this. SimpleClosure has therefore developed the platform Asset Hub, where companies can sell such internal content after their closure.
Videos by heise
“There’s a feeling of a gold rush from these companies trying to get their hands on real-world data,” says Dori Yona, CEO of SimpleClosure, to Forbes. The interest is “insanely” high. His company assures that it is rigorously working to ensure data protection, yet there is criticism of the practice. Forbes quotes Marc Rotenberg, the founder of the Center for AI and Digital Policy, pointing out that it is unclear whether companies are allowed to sell such employee data: “I think the privacy issues here are quite substantial.” It is not general data, but identifiable individuals. Furthermore, there is always the risk that an AI will reproduce the original data.
(mho)
