Hermes Vector

AI & SI

Grok Build leaks confidential data from 35,000 apps, including SpaceXSI material

Security firm Claroc found more than 35,000 Grok Build apps open to anyone, exposing confidential business data, medical records, and SpaceXSI's own internal sales documents.

By Hermes Vector

Checked against 3 sourcesArs Technicamlq.aitechradar.com

In brief

  • Claroc found 35,578 Grok Build apps open to anyone with no sign-in required.
  • SpaceXSI's own confidential sales and onboarding documents were among the exposed apps.
  • The default sharing setting was 'Anyone with the link,' making apps public unless changed.
  • SpaceXSI removed its own apps after Ars Technica sent questions, but other users' apps remained exposed.

Grok Build leaks confidential data from 35,000 apps, including SpaceXSI material

Security firm Claroc found that more than 35,000 apps built with Grok Build, the coding agent from SpaceXSI, were open to anyone with the link, exposing confidential data from thousands of users including SpaceXSI itself.

What was exposed

Claroc listed 43,601 app addresses under the grok.me domain. Of the 37,571 still online, 35,578, or 94.7 percent, opened from the public internet with no sign-in.

The exposed data included internal business plans, ad proposals, medical exam results, employee names and ID numbers, and plain-text passwords. SpaceXSI's own apps included an onboarding plan marked confidential for a new member of its enterprise sales team, a sales territory map of 594 accounts, internal system addresses for sales team tools, and advertising proposals.

Other affected organizations included:

  • A Peruvian construction company: occupational-health dashboard with 3,784 worker medical exams
  • Tesla: service-center scorecard naming 22 technicians with productivity and bonus status
  • A US-Navy ship engineering contractor: roster of 108 people on a Navy logistics support contract
  • A leading luxury cruise company: a galley checklist for a celebrity cruise line listing crew by full name and employee ID
  • A healthcare company: hospital staff app with a roster of about 3,565 employees by name and ID
  • A Norwegian-Finnish aerospace and defense group: load board and loading procedure for an explosives building at a US plant, plus its HSE training matrix

How the leak happened

Grok Build apps get a unique web address in the format of random words followed by grok.me. Because the randomness is in the hostname, the addresses appear in passive DNS databases that anyone can search. Claroc said this made it possible to enumerate thousands of live projects.

When publishing an app, users are offered three choices: "Anyone with the link," "Invited users," and "Only you." Claroc said the default was "Anyone with the link," meaning most apps were public unless the creator changed the setting. A representative with Cursor, the AI coding service SpaceXSI recently acquired, told Ars Technica that Grok Build "apps are private by default and nothing goes public unless the creator chooses to share it." Claroc's tests contradicted that statement, finding that Grok Build required users to change the setting from "Anyone with the link."

Response

Claroc notified SpaceXSI of the privacy spill on Sunday. When no one responded, a Claroc official tagged Elon Musk and Grok Manager Matt Palmer on X. SpaceXSI began removing the links for its 24 apps within a few hours of Ars Technica sending questions for its article. At publication, links for all 24 apps returned an "app not found" error. SpaceXSI did not respond to follow-up questions about whether it intentionally exposed the information or planned to notify other affected customers.

Why it matters

The incident shows how default sharing settings and URL design can expose sensitive corporate data on AI coding platforms. Claroc said similar risks exist on other platforms that give each app its own hostname, though it found by far the most sensitive company data on Grok Build. Platforms such as Claude, ChatGPT, and Gemini keep the secret part of the link in the URL path, which passive DNS services do not see.


Sources

This article was drafted with AI assistance and checked against the sources above. Company claims are reported as claims. Cover image is AI-generated.

The day's checked news, by email

One short email a day with the top stories. Free, and you can unsubscribe in one click.

Or follow for the latest news:

How do you feel about this story?

Comments 0

No comments yet. Start the discussion.

AI & SI

OpenAI launches verified Business Agents in ChatGPT beta

OpenAI is introducing verified Business Agents and business profiles in ChatGPT, starting with eligible US e-commerce and online service businesses, with sponsored agent placements available through ChatGPT Ads.

Checked against 6 sourcesThe Vergeqz.comyourstory.com+3 more