Security researchers have uncovered a major vulnerability in Grok Build, an AI-powered tool for creating applications. More than 35,000 apps were accessible without authentication. Some contained confidential corporate documents, medical records, or employee passwords. Companies associated with Elon Musk were also among those affected.
Sensitive Data Exposed at Elon Musk’s Companies
The issue was discovered by cybersecurity firm Claroc, which analyzed applications created using Grok Build. The tool allows users to develop functional applications through simple text prompts.
Researchers examined more than 43,000 web addresses. Of the 37,571 active applications identified, 35,578, or 94.7%, were accessible to anyone without authentication.
Most were ordinary applications containing no sensitive information. However, researchers estimated that approximately 650 to 800 projects may have exposed private data.
The most serious cases involved patient medical records, employee lists containing thousands of names, and corporate databases. One application even exposed passwords for 18 user accounts in plain text.
The security issue also affected Grok’s operator and the social media platform X. Researchers identified 24 publicly accessible internal applications belonging to these companies.
These contained business plans, advertising proposals, and a database of 594 strategic customer accounts. Additional exposed information originated from at least 30 identified organizations.
Weak Security Exposed Thousands of Applications
The main cause was not a traditional cyberattack but the way applications were published. Grok assigns unique web addresses that are supposed to be difficult to guess.
However, these addresses could be discovered through publicly searchable DNS databases. Simply knowing an application’s URL was often enough to open it and access stored information.
Researchers also found that their test account offered public access by default when publishing applications. In another test application, an unauthorized user was able to modify and delete customer records.
The company operating Grok stated that applications remain private until their creators publish them. Following the disclosure, it removed its 24 affected internal applications.
However, according to researchers, other publicly accessible projects remained exposed. There is currently no confirmed evidence that the exposed data has been misused.
The incident highlights growing security risks associated with AI-generated applications. Users should carefully review access permissions and avoid uploading sensitive information to such tools without adequate protection.
