AI Labs Target Defunct Startup Data for Agent Training

AI laboratories are increasingly targeting intellectual property from defunct startups to train autonomous agents on complex professional tasks, according to industry observers. This emerging practice involves acquiring code repositories, Slack histories, Figma designs, and internal documentation from failed ventures to simulate real-world work environments. Proponents argue this data provides essential context for debugging and workflow automation that synthetic datasets cannot replicate. However, legal experts warn that bankruptcy proceedings rarely address AI training rights explicitly, creating significant liability regarding employee privacy and proprietary information. The trend highlights a critical gap in current insolvency frameworks as AI firms seek high-fidelity operational data. While no specific acquisition deals have been publicly confirmed, the strategic demand for authentic enterprise workflows suggests this will become a standard data sourcing channel. Stakeholders now face urgent questions about consent and valuation of digital remains.