Elon Musk’s satellite internet venture, Starlink, has quietly updated its global privacy policy to allow the use of customer data for training artificial intelligence, sparking fresh concerns over digital surveillance and personal privacy.
The policy shift, which came into effect on 15 January 2026, means that unless users manually “opt out,” their information may be used to develop machine learning models.
Furthermore, this data could be shared with various service providers and “third-party collaborators,” though the company has yet to provide specific details on who these partners are, Reuters reported.
Analysts suggest the change is timed to support Mr Musk’s broader corporate ambitions. For instance, SpaceX, Starlink’s parent company, is currently in talks to merge with xAI—Mr Musk’s dedicated artificial intelligence firm—which was recently valued at $230 billion.
Consequently, the merger would provide xAI with a massive new reservoir of real-world data from Starlink’s 9,000 satellites and more than 9 million users.
Starlink’s internal privacy documentation confirms that the company currently harvests an extensive range of personal information from its global user base.
This data collection includes precise location metrics and user IP addresses, which are used to track the physical and digital footprint of customers.
Furthermore, the service maintains records of credit card information and personal contact details, as well as more sensitive communication data.
This latter category is particularly broad, encompassing audio recordings, visual information, and any files shared across the network, all of which now potentially fall under the company’s new artificial intelligence training protocols.
| Data Category | Specific Examples Collected | Potential AI Use Case |
| Connectivity Data | IP addresses, latency, signal strength | Network optimization and predictive routing |
| Personal Identity | Credit cards, contact info, billing history | Fraud detection and consumer behavior modeling |
| Sensitive Content | Audio, video files, communication logs | Natural Language Processing (NLP) and voice recognition |
| Geospatial Data | GPS location, movement patterns | Geographic demand forecasting and mapping |
While the revised policy does not specify which categories will be fed into AI models, it marks a significant departure from previous versions of the policy reviewed as recently as November, which made no mention of AI training.
The update comes at a pivotal moment for SpaceX as it prepares for a blockbuster Initial Public Offering (IPO) later in 2026.
Financial experts predict that the listing could push the company’s valuation north of $1 trillion.
By integrating Starlink’s data with xAI’s “Grok” chatbot and other AI services, Mr Musk aims to create what some describe as a “vertically integrated innovation engine.”
However, for now, the burden remains on the consumer to navigate their account settings and uncheck the box if they wish to keep their data out of the training loop.
