SpaceX's satellite internet service Starlink revised its Global Privacy Policy on January 15 to state that, unless users opt out, Starlink-collected data may be used to train the company's machine learning or artificial intelligence models and shared with service providers and third-party collaborators.
The policy change was not present in an archived November version of Starlink's policy, according to a review of the prior text. The newer language does not specify precisely which categories of data would be employed for AI training, leaving that detail unspecified in the updated document.
Starlink gathers a wide array of customer information under its privacy framework. That collection includes location data, credit card details, contact information and user IP addresses. The policy also identifies a category described as "communication data," which encompasses audio and visual content, files shared by users and "inferences we may make from other personal information we collect." The updated policy indicates this data could be used for the company’s machine learning or AI models unless an individual opts out.
Industry observers and consumer advocates have flagged concerns about the move. Using personal communication data to train AI, they say, expands the potential for surveillance and opens new avenues for misuse unless tightly constrained. "It certainly raises my eyebrow and would make me concerned if I was a Starlink user," said Anupam Chander, a technology law professor at Georgetown University. "Often there’s perfectly legitimate uses of your data, but it doesn’t have a clear limit to what kind of uses it will be put to."
The policy revision comes as SpaceX is reportedly in discussions to merge with the AI firm xAI ahead of a planned SpaceX initial public offering later this year. Those talks, if they proceed, would combine Starlink's data trove with xAI's model development efforts. xAI was most recently valued at $230 billion following a recent funding round and is developing the Grok large language model chatbot. The company also owns the social media platform X.
Starlink operates a constellation of more than 9,000 satellites and currently serves over 9 million users. The scale of the network and its user base mean the potential data available for model training - including communication content - is substantial, according to the updated policy language.
SpaceX did not respond to a request for comment on the policy change.
Context and implications
The updated policy introduces a new permissive posture toward the use of customer information for AI development while leaving open questions about the precise data elements that would be incorporated and the safeguards that would govern access and sharing. The change intersects with corporate strategy - including reported merger discussions between SpaceX and xAI and a planned IPO for SpaceX - and has prompted scrutiny from privacy experts about the breadth and limits of permissible uses.