
The information displayed in the AIM should not be reported as representing the official views of the OECD or of its member countries.
Defunct startups are selling internal communications—such as Slack messages, emails, and documents—to AI companies for use in training advanced models. This practice, facilitated by firms like SimpleClosure, has raised significant privacy concerns, as sensitive employee data is repurposed without clear consent, potentially violating privacy rights.[AI generated]
Why's our monitor labelling this an incident or hazard?
The article explicitly involves AI systems, specifically AI agents trained on real-world workplace data derived from defunct companies' internal communications. The use of this data for AI training is ongoing and expanding, with companies actively buying and selling such data. Although no direct harm (such as a data leak or privacy violation) is reported as having occurred yet, the article details substantial privacy concerns and technical difficulties in anonymizing data, which could plausibly lead to violations of privacy rights and other harms if AI models memorize and reveal sensitive information. This fits the definition of an AI Hazard, as the development and use of AI systems with this data could plausibly lead to an AI Incident involving privacy violations. The article does not describe a realized harm or incident, nor is it primarily about responses or governance measures, so it is not Complementary Information. It is not unrelated because AI systems and their impacts are central to the narrative.[AI generated]