As defunct companies' digital archives change hands for training AI, privacy advocates warn of substantial risks, despite rising commercial incentives among AI developers for real workplace data.
As the shutdown market around failed startups becomes more organised, some founders are now turning their company’s private digital trail into a final asset. Emails, Slack messages, Jira records and other workplace files are being packaged up and sold to artificial intelligence developers, with some deals reportedly worth tens of thousands of dollars and, in some cases, far more.
Forbes reported that AI labs are buying these archives to build what the industry calls reinforcement learning gyms: simulated workplace environments made from real company data, where agents can be trained to handle tasks that mirror everyday office work. That kind of material has become increasingly valuable as AI systems move beyond scraping public websites and towards training on data that captures how organisations actually operate.
SimpleClosure, a company that helps startups wind down by dealing with payroll, taxes and investor settlements, has moved into the market with a product called Asset Hub. According to the company, it helps sellers identify what data may be licensable, estimate its worth and strip out personally identifiable information before a deal is completed.
The commercial incentive is clear, but so are the privacy concerns. Forbes said SimpleClosure has handled nearly 100 such transactions over the past year, with payouts ranging from $10,000 to $100,000 per company. Marc Rotenberg, founder of the Centre for AI and Digital Policy, told Forbes that the privacy issues are substantial, arguing that employee messages are not generic training material but identifiable records of real people. His group has also urged the US Senate Commerce Committee to press the Federal Trade Commission for tighter oversight of AI-related businesses.
Source Reference Map
Inspired by headline at: [1]
Sources by paragraph: - Paragraph 1: [2], [3] - Paragraph 2: [3], [4] - Paragraph 3: [4] - Paragraph 4: [2], [3], [4], [5]
Source: Noah Wire Services
Verification / Sources
- https://gizmodo.com/failed-companies-are-selling-old-slack-chats-and-email-archives-to-train-ai-2000747916 - Please view link - unable to able to access data
- https://gizmodo.com/failed-companies-are-selling-old-slack-chats-and-email-archives-to-train-ai-2000747916 - This article discusses how defunct startups are selling their internal communications, including emails and Slack messages, to train AI models. Companies like SimpleClosure are facilitating these transactions, allowing startups to monetize their data before shutting down. The article highlights the lucrative nature of this data for AI training and raises privacy concerns regarding the inclusion of identifiable employee information in these datasets.
- https://www.forbes.com/sites/annatong/2025/04/16/ai-labs-are-buying-slack-jira-and-email-archives-from-defunct-startups-to-build-reinforcement-learning-gyms-and-train-ai-agents-in-simulated-workplaces/ - Forbes reports that AI labs are purchasing Slack, Jira, and email archives from defunct startups to create 'reinforcement learning gyms.' These simulated environments use real-world company data to train AI agents in workplace tasks, reflecting the increasing value of such data for AI development.
- https://www.simpleclosure.org/ - SimpleClosure is a company that assists startups in shutting down by handling tasks like payroll, taxes, and investor settlements. They offer a platform called Asset Hub, which helps startups monetize their data, including internal communications like Slack messages and emails, by licensing them to AI labs for training purposes.
- https://arstechnica.com/tech-policy/2024/05/slack-defends-default-opt-in-for-ai-training-on-chats-amid-user-outrage/ - Ars Technica reports on Slack's policy of using customer data, including messages and files, to train its AI models by default. Users have expressed concerns over this practice, leading Slack to consider updating its privacy policies to provide clearer information on how user data is utilized for AI training.
- https://www.engadget.com/yuck-slack-has-been-scanning-your-messages-to-train-its-ai-models-181918245.html/ - Engadget highlights Slack's practice of scanning user messages, files, and other content to train its machine learning models without explicit permission. The article discusses the opt-out process, which requires users to have their workspace owner contact Slack directly, raising concerns about user privacy and data usage.
- https://www.bespacific.com/slack-has-been-using-data-from-your-chats-to-train-its-machine-learning-models/ - Bespacific reports on Slack's use of customer data, including messages and files, to train its machine learning models. The article details the opt-out process, which necessitates workspace owners contacting Slack to stop data usage, and discusses the implications for user privacy and data control.
Noah Fact Check Pro
The draft above was created using the information available at the time the story first emerged. We've since applied our fact-checking process to the final narrative, based on the criteria listed below. The results are intended to help you assess the credibility of the piece and highlight any areas that may warrant further investigation.
Freshness check
Score: 8
Notes: The article was published on April 17, 2026, and reports on a Forbes article from April 16, 2026. (forbes.com) The information appears to be current and original, with no evidence of prior publication or recycled content. However, the reliance on a single source (Forbes) for the primary information raises concerns about source independence.
Quotes check
Score: 7
Notes: The article includes direct quotes from Marc Rotenberg, founder of the Center for AI and Digital Policy, and Dori Yona, CEO of SimpleClosure. (forbes.com) While these quotes are attributed, they cannot be independently verified through other sources, which diminishes their reliability.
Source reliability
Score: 6
Notes: The primary source is Forbes, a reputable news organisation. However, the article heavily relies on a single source for its information, which raises concerns about the diversity and independence of the sources used. Additionally, the article includes a link to SimpleClosure's website, which may be considered promotional content and not an independent verification source.
Plausibility check
Score: 7
Notes: The concept of defunct startups selling their internal communications to AI labs for training purposes is plausible and aligns with current trends in AI development. However, the article lacks specific details about the companies involved, the exact nature of the data sold, and the terms of these transactions, which makes it difficult to fully assess the accuracy of the claims.
Overall assessment
Verdict (FAIL, OPEN, PASS): FAIL
Confidence (LOW, MEDIUM, HIGH): MEDIUM
Summary: The article presents plausible information about defunct startups selling their internal communications to AI labs for training purposes. However, it heavily relies on a single source (Forbes) and includes unverifiable quotes, which raises concerns about source independence and the ability to independently verify the claims. The lack of specific details about the companies involved and the terms of these transactions further diminishes the article's reliability. Given these issues, the content cannot be fully verified, and publishing is not covered under our indemnity.