On November 10, 2026, security firm Claroc reported that 35,578 of 37,571 Grok apps still online in its test opened without sign-in. That is 94.7% of the apps in that tested set. Claroc also found exposed business and personal information, but estimated that sensitive material appeared in about one in ten open apps—not all of them. (Claroc’s report)
What the numbers cover
Claroc identified 43,601 names under grok.me and tried 43,551 app addresses. It found 37,571 apps still online; 35,578 opened without sign-in, while 1,993 returned an HTTP 401 response. The identified names, tried addresses and online apps are separate stages of Claroc’s count.
Claroc estimated that about one in ten open apps contained sensitive, internal-business or personal information, and roughly 2% contained material it rated high or critical. The estimates drew on automated flagging, manual review and sample checks. They are not counts of every app manually inspected.
Among the examples Claroc described were 24 accessible apps created by SpaceXSI and X staff, including a confidential onboarding plan and a sales-territory map covering 594 accounts. Claroc also reported a contractor timesheet script containing plaintext passwords for 18 employee accounts.
Why a random hostname could be found
Passive DNS is a record of domain-name lookups collected over time. When someone looks up an app’s individual hostname, that hostname can enter passive-DNS indexes and become searchable, even if it contains random words. A secret placed in the path of a URL—the part after the domain name—is not visible to DNS, though search engines and link scanners can still expose it, Claroc said.
| Link structure | What DNS can see | Other exposure described by Claroc |
App-specific hostname, such as a Grok app under grok.me | The queried hostname | Passive-DNS indexes can make the hostname searchable. |
| Secret identifier in a URL path | The hostname, but not the path | Search engines or link scanners can still expose the link. |
A difficult-to-guess address can make discovery harder, but it does not replace controlling who is allowed to open an app.
What Claroc’s tests demonstrated
Claroc tested two apps it had created with invented data. In one, an unauthenticated visitor could add a contact. In the other, a Google account that had not been invited could view, change and delete customer records.
Those write and access-control tests were confined to Claroc’s own apps. Claroc said it did not sign in to other people’s apps, submit their forms or modify their data.
Sharing settings and proposed changes
In Claroc’s paid test account, “Anyone with the link” was selected in the publishing panel. Separately, Claroc reported that SpaceXSI told a reporter employee accounts defaulted to “Only you.” Those statements refer to different account contexts.
Claroc said it notified SpaceXSI on November 8, 2026, and reported 22 additional apps made by SpaceXSI and X staff on November 9. It reported that SpaceXSI took down the 24 staff-created apps after its report, published November 10.
Claroc proposed placing shared apps on hard-to-guess paths under a single host, allowing access restrictions by organization email domain, making “Only you” the default, and showing who can access an app before publication. It also recommended warning users when an app has a database but no sign-in requirement.
