Over 35,000 Grok‑Built Apps Exposed Online, Some Leaking Confidential Data and Passwords
More than 35,000 Grok‑built apps are publicly reachable, with 24 containing internal confidential data from X and SpaceXSI and many others exposing sensitive information from Fortune 500 firms. The exposure stems from Grok’s default “Anyone with the link” setting.
More than 35,000 AI‑generated applications built on Grok, the low‑code app builder that lets users create and share custom tools, are reachable on the public internet without any sign‑in requirement. The situation surfaced after security firm Claroc scanned the web and found that the platform’s default sharing setting leaves apps open to anyone who has the link, unless the creator deliberately tightens access.
Scale of the exposure
Claroc’s internet‑wide sweep uncovered 37,571 Grok‑built apps in total. Of those, 35,578 were configured to allow unrestricted access, meaning anyone could view the app simply by knowing the URL. This massive surface area is a direct result of Grok’s default option of “Anyone with the link” for paid accounts, a setting that remains active unless the user changes it.
Confidential data from internal users
Among the publicly visible apps, 24 were created by staff at X and SpaceXSI. These internal apps stored a range of confidential material, including: - A sales onboarding plan for a new hire. - A sales‑territory map covering 594 accounts. - An advertising proposal prepared for a U.S. federal agency, marked confidential.
SpaceXSI (formerly known as xAI) has taken the step of removing all 24 staff‑built apps, but it has not yet removed apps that were created by other, non‑staff users.

Photo by Christina Morillo on Pexels
Data from external organisations
The remaining open apps also exposed internal information from 30 other organisations that Claroc could identify. Of those, 26 employ more than 1,000 staff members and six belong to the Fortune 500 list. Roughly 2 % of the open apps contained material that Claroc rated as high or critical severity. Sample data items that appeared in these apps include: - A searchable employee list from a hospital group containing over 3,500 names. - An intensive‑care unit (ICU) briefing for a single patient. - A live shift board used by a fire department. - A contractor timesheet that stored every employee’s password in plain‑text.
These examples illustrate that the exposure is not limited to trivial artefacts, it reaches into highly sensitive operational domains.
What organisations should do now
The breach highlights a straightforward but often overlooked risk: default sharing settings on cloud‑based app builders can make internal data globally visible. The following pragmatic steps can reduce exposure quickly:
- Audit existing Grok apps, Use passive DNS queries or simple URL‑scanning tools to locate any Grok‑built apps that resolve to public IPs.
- Restrict access, Change each app’s sharing setting from “Anyone with the link” to either “Only you” or “Invited users”. This can be done from the app’s settings page.
- Review content, Open each app and verify that no internal documents, credentials, or operational data are present. If any such material is found, remove it or move it to a secured repository.
- Unpublish or delete, For apps that were never intended for public consumption, unpublish them from the Grok marketplace or delete them entirely.
- Establish governance, Implement a company‑wide policy that mandates a security review before any AI‑generated app is shared outside the organisation. The policy should cover who can create apps, which sharing settings are allowed, and how often audits are performed.
- Leverage expert help, Organisations that lack in‑house expertise can work with specialists to audit and harden their AI‑built applications. Stratum’s AI consultancy service, for example, offers a structured review of Grok apps and can help put safeguards in place. (Stratum AI consultancy)

Photo by panumas nikhomkhai on Pexels
Longer‑term considerations
While the immediate actions above aim to plug the most obvious gaps, companies should also think about how they integrate AI‑generated tools into their broader security posture. Key questions include: - How are credentials managed within AI‑built apps? - Are logs and audit trails available for app‑level activity? - Does the organisation have a clear incident‑response plan for accidental data exposure from AI tools?
Answering these questions now can prevent similar incidents as AI‑assisted development becomes more prevalent.
Bottom line
The discovery of over 35,000 publicly accessible Grok‑built apps, some leaking employee lists, ICU briefings and even plain‑text passwords, is a stark reminder that ease of sharing can become a liability. By auditing existing apps, tightening sharing settings, and instituting clear governance, businesses can protect sensitive data while still benefitting from the productivity gains of AI‑driven development.