Elon Musk's SpaceXAI is internally discussing the acquisition of customer and operational data from failed startups to train its Grok AI model. This approach aims to secure vast datasets cheaply, similar to Google's past purchase of Spirit Airlines' business records. The move raises privacy concerns, as highlighted by an ongoing court battle over employee data.

The pursuit of vast datasets for AI training, even from defunct companies, highlights the escalating costs and ethical considerations in AI development, potentially impacting user privacy and data rights.
SpaceXAI, the artificial intelligence division of Elon Musk's SpaceX, is reportedly holding informal internal discussions about purchasing customer and operational data from defunct startups. The aim is to acquire large datasets at a lower cost for training the Grok AI model, a strategy that mirrors Google's $10 million acquisition of Spirit Airlines' business records in a bankruptcy auction.
AI models like Grok learn from training data, and acquiring such data from companies that have ceased operations is seen as a potentially cost-effective method compared to scraping web pages or licensing data from active companies. The Spirit Airlines acquisition included approximately 100 million emails and 500 million Microsoft Teams messages, along with decades of employee files.
This practice has raised privacy concerns. A flight attendants' union has objected in bankruptcy court to the use of Spirit Airlines' employee data, arguing that even with de-identification, individuals could be identified. The court case is ongoing. Bankruptcy proceedings treat customer databases similarly to other assets, auctioning them to the highest bidder without direct consent from the original customers.
In August, Musk informed SpaceX employees that the company would also utilize its own internal records for Grok's training, reportedly stating that the model would "inherit your thoughts." The extent of which employee data categories would be used remains unspecified.