As distressed startups liquidate internal emails and chat logs to AI firms, experts warn of privacy breaches and legal loopholes amid technical limitations in anonymising data.
As startups collapse, a more unsettling asset is being liquidated alongside office furniture and software licences: years of internal emails, chat logs and other workplace communications. According to the Clubic report, some distressed companies are now selling those records to firms that want them for AI training, turning mundane business correspondence into a commercial dataset.
That practice sits on firmer legal ground than many employees might expect. Slack says on its own trust and terms pages that the customer, not the individual user, owns and controls the content of a workspace, including messages and files, while administrators can manage retention, export and deletion settings. In other words, the employer typically holds the keys to the archive, even if the material was written by staff during the course of their jobs.
But the fact that a company controls the data does not settle the privacy question. Marc Rotenberg of the Center for AI and Digital Policy argued that these records can contain personally identifiable information, not merely abstract business text, and that assignment of workplace IP rights does not automatically justify resale of internal communications to a third party. The organisation has urged US lawmakers to press the Federal Trade Commission to scrutinise the market more closely.
The technical safeguards are also far from watertight. OpenAI and Google researchers showed in a 2020 study that large language models can memorise portions of their training data and later reveal them with the right prompts, which undercuts easy claims that anonymisation alone makes the process safe. That concern is one reason data buyers tend to stress privacy controls so heavily, even as critics argue that stripping personal traces from a career’s worth of messages is not something software can reliably do. In Clubic’s reporting, former cielo24 chief executive Shanna Johnson said she received “des centaines de milliers de dollars” for 13 years of internal company data, while Bobby Samuels of Protege, which handles regulatory-risk data, said there is no technical shortcut for erasing a person’s footprint from such a dataset.
Source Reference Map
Inspired by headline at: [1]
Sources by paragraph:
Source: Noah Wire Services
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 18, 2026, and reports on recent developments regarding defunct startups selling internal communications for AI training. Similar reports from April 16, 2026, in Forbes and April 17, 2026, in Gizmodo indicate that the narrative is fresh and original. However, the Clubic article does not provide specific publication dates for the sources it references, which raises concerns about the freshness of the information. ([forbes.com](https://www.forbes.com/sites/annatong/2026/04/16/ais-new-training-data-your-old-work-slacks-and-emails//?utm_source=openai))
Quotes check
Score:
7
Notes:
The article includes a quote from Marc Rotenberg of the Center for AI and Digital Policy, stating that “employee privacy remains a key concern, particularly because people have become so dependent on these new internal messaging tools like Slack.” A search for this exact quote reveals it was used in a Forbes article published on April 16, 2026. ([forbes.com](https://www.forbes.com/sites/annatong/2026/04/16/ais-new-training-data-your-old-work-slacks-and-emails//?utm_source=openai)) This suggests the quote may have been reused, raising concerns about originality. Additionally, the Clubic article does not provide direct links to the sources of its quotes, making independent verification challenging.
Source reliability
Score:
6
Notes:
Clubic is a French technology news website. While it is a known publication, it is not as widely recognised in the English-speaking world. The article references sources such as Forbes and Gizmodo, which are reputable English-language publications. However, the lack of direct links to these sources in the Clubic article makes it difficult to assess the reliability of the information presented.
Plausibility check
Score:
9
Notes:
The practice of defunct startups selling internal communications for AI training aligns with recent trends in AI development, where companies seek diverse datasets for training models. Reports from Forbes and Gizmodo corroborate this practice, indicating that the claims are plausible. However, the Clubic article does not provide specific examples or detailed information, which limits the ability to fully assess the plausibility of the claims.
Overall assessment
Verdict (FAIL, OPEN, PASS): FAIL
Confidence (LOW, MEDIUM, HIGH): MEDIUM
Summary:
The article presents a plausible and timely topic, supported by reports from reputable sources. However, the lack of direct citations, potential reuse of quotes, and challenges in verifying the freshness and originality of the content raise significant concerns. Given these issues, the content does not meet the necessary standards for publication under our editorial indemnity.
