Editorial illustration for Grok Build coding agent leaks confidential data, some belonging to SpaceXSI
Grok Build Leaks Confidential Data From SpaceX Apps
Grok Build coding agent leaks confidential data, some belonging to SpaceXSI
Grok Build, the vibe-coding tool xAI's SpaceXSI unit sells to developers, has been handing out confidential data to anyone who knows where to look. Security firm Claroc found that thousands of apps built on the platform are exposing internal documents, medical records, and business files through share links that were supposed to be randomized and private.
The setup was simple enough in theory. Grok Build lets users generate a shareable URL for their app, and that URL is meant to be unguessable unless someone already has it. But a design choice in how those links get generated means outsiders can enumerate live projects on the platform almost at will. Click one, and in most cases you get the whole app, data included.
Among the exposed material: internal business plans, advertising proposals, medical exam results, galley checklists from a cruise line, and employee names paired with ID numbers. The irony, according to Claroc's Tuesday writeup, is that some of the leakiest apps belong to SpaceXSI itself, the same company that built and sells Grok Build in the first place.
The information is revealed through links allowing Grok Build users to share their apps with others. Grok takes steps to randomize the URLs so only people who know the exact URL can access the data stored there. A design decision by developers of the platform, however, completely torpedoes that objective and makes it possible for anyone to enumerate thousands of live Grok Build projects.
Why this matters
This is a URL-as-security-boundary failure, the oldest mistake in the book, and it's happening at a company owned by the same man who built the car with the autopilot problem and the social network with the verification problem. Randomized links are not access control. If you can guess or scrape a URL, you don't need a password, and Claroc's researchers apparently didn't need much more than that to pull internal business plans and medical records out of a tool meant for quick app prototyping.
For developers building on vibe-coding platforms, the lesson is blunt: "shareable by link" almost always means "discoverable by anyone," and you should assume default sharing settings leak until proven otherwise. For founders evaluating these tools for internal workflows, this is a reminder that convenience features ship faster than the access logic behind them gets audited. And for xAI, the embarrassment is specific: its own coding agent exposed data tied to SpaceX, the company's biggest customer for credibility.
Build fast, audit faster, or expect your own platform to out you.
Related Reading
Research & Benchmarks Microsoft AI Chief: "If It Isn't Safe We Shouldn't Build It
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Related Topic Anthropic sues Trump as xAI restarts AI coder push, Iran releases AI fakes
Open Source
Related Topic Anand: Gmail black-box limits AI reverse engineering of proprietary software
AI Tools & Apps
From Archives DeepMind Reveals Gemini Robotics 2.5 and Open Safety Benchmark ASIMOV-Jailbreak
From Archives AI Agents Test Zero-Trust Limits as Single PEP Scaling Falls Short
From Archives Coforge Joins SpaceX Network to Develop Industry-Specific Grok Apps
Business & Startups
From Archives SpaceXSI's Grok Bot AI agent makes phone calls; Tencent releases Hy4 model
Open Source
Common Questions Answered
What security vulnerability did Claroc find in Grok Build's shareable URLs?
Claroc discovered that Grok Build's supposedly randomized and private share links could be easily enumerated, allowing attackers to access thousands of live projects without authorization. The platform's design decision to use randomized URLs as the primary security mechanism failed because the URLs were guessable or scrapable, exposing internal documents, medical records, and business files to anyone who knew where to look.
How does the Grok Build URL-as-security-boundary failure work?
Grok Build relies on randomized URLs as its access control mechanism, assuming that unguessable links would keep shared apps private. However, this approach is fundamentally flawed because if someone can guess or scrape a URL, they gain full access without needing a password or any other authentication method, making randomization alone insufficient for security.
What types of confidential data were exposed through Grok Build's vulnerable share links?
Security researchers found that thousands of apps built on Grok Build were exposing internal documents, medical records, and business files through the insecure share links. This included sensitive information from SpaceX and other organizations using the platform, demonstrating the serious real-world impact of the vulnerability.
Why is Grok Build's security flaw described as 'the oldest mistake in the book'?
The article characterizes this as a URL-as-security-boundary failure, which is a well-known and repeatedly-made security mistake where developers rely on obscure URLs instead of proper authentication and access control mechanisms. This fundamental design flaw has been exploited for decades and represents a basic misunderstanding of security principles.
Further Reading
- Grok Build coding agent leaks oodles of confidential info, some belonging to SpaceXSI - Ars Technica
- GhostShares 2: Grok Apps Exposed xAI’s Business Secrets and People’s Private Data - Claroc
- Grok Build exposed confidential data, including SpaceXSI material - Mezha
- Grok Build left 35,000 apps open to anyone, including SpaceX's own confidential sales material - MLQ.ai
- Grok Build’s App Sharing Puts Private Data at Risk, Researchers Say - TeslaNorth
