SpaceX (SPCX.US)'s AI division sets its sights on "bankruptcy data"! Plans to acquire customer and operational information of struggling startups to ramp up Grok model training.

Title context: SpaceX (SPCX.US)'s AI division sets its sights on "bankruptcy data"! Plans to acquire customer and operational information of struggling startups to ramp up Grok model training. Text: According to people familiar with the matter, SpaceX (SPCX.US), owned by Musk, has internally discussed purchasing customer and operational data from struggling or already failed startups, hoping to obtain more high-quality data at a relatively low cost to improve the performance of its artificial intelligence models. The people said the discussions are currently mainly taking place within SpaceXAI, SpaceX's AI division, and remain at an informal stage, and may ultimately not result in any deal. SpaceX did not respond to a request for comment. This approach is quite similar to a previous move by Alphabet Inc. Class C. Earlier this year, after U.S. budget airline Spirit Airlines ceased operations, Alphabet Inc. Class C offered $10 million to purchase its commercial data for AI training. However, the deal also raised concerns about data privacy, including questions raised by some former Spirit Airlines flight attendants. Intensifying AI competition drives SpaceXAI to seek more external high-quality data SpaceXAI was formerly known as xAI. As Musk's space, satellite communications and artificial intelligence businesses become further integrated, the company is competing with AI companies such as Anthropic and OpenAI, and seeking to attract more enterprise customers. Like other AI companies, one of SpaceXAI's key tasks is to find high-quality training data that can improve model performance across different tasks. According to people familiar with the matter, the company is particularly focused on enterprise operational data and customer information, and plans to use such external data to train models including Grok. If ultimately implemented, this would mean a certain shift in SpaceXAI's data strategy. In the past, the company mainly relied on data from Musk's social platform X, as well as an internal team of professionals known as "AI tutors," to train and optimize its artificial intelligence models. These AI tutors help engineers improve the model's performance in specialized fields ranging from finance and science to humor. SpaceXAI now considering bringing in more external datasets reflects that, as AI model capabilities continue to improve, relying solely on internal data sources may no longer be enough to fully meet model training and optimization needs. For leading AI companies, being able to legally obtain data that is specialized, authentic and structured is becoming an important resource for improving model performance. Reorganizing AI training team to accelerate improvement of Grok training system At the same time, the data team responsible for model training within SpaceXAI has also recently undergone a series of personnel and strategic adjustments. In June this year, SpaceXAI temporarily paused hiring AI tutors responsible for training Grok, and subsequently adjusted the team's leadership. Jack Garabedian, who had long worked at SpaceX's satellite internet business Starlink, recently took over the team, replacing young engineer Diego Pasini. People familiar with the matter said that since taking office, Garabedian has been trying to improve conditions within the team, including establishing clearer workflows, meeting mechanisms and AI training data goals. An internal SpaceXAI communication seen by the media showed that the data team's progress in July included advancing the development of a new AI model and a coding agent. The communication emphasized that without the team's daily foundational work such as labeling, preference alignment and verification, it would be difficult for these models to eventually become mature products. This also highlights that, beyond computing power and model architecture, data processing and human feedback remain important components of AI model training. External procurement and internal data in parallel: Musk plans to have Grok learn "all SpaceX information" Although SpaceXAI is studying ways to obtain more external data sources, Musk's vast business empire itself remains an important data source for Grok, including information generated by SpaceX employees. Musk recently said at an internal SpaceX meeting that the company plans to use all information within SpaceX to train Grok, and directly told employees: "It will also be trained on your data." This statement means that SpaceXAI's future data strategy may adopt a parallel "internal + external" model: on the one hand, it will continue to use data generated by Musk-owned companies and platforms such as SpaceX and X; on the other hand, it will further expand the breadth and specialization of training data by purchasing external datasets. It is worth noting that although acquiring data from struggling companies or failed businesses may provide a relatively low-cost data source, the use of customer information may also involve issues such as privacy, authorization and the scope of data use. The controversy previously triggered by Alphabet Inc. Class C's bid for Spirit Airlines data has already shown the potential compliance and privacy risks of such transactions. As companies such as OpenAI, Anthropic and SpaceXAI compete over model performance and enterprise customers, high-quality, specialized data is becoming another key resource in AI competition after chips, computing power and electricity. SpaceXAI's consideration of purchasing operational and customer data from struggling startups also shows that the competition among leading AI companies over training data is further extending into the field of private enterprise data.