AI-Coded Apps from Lovable, Replit, Base44 Expose Corporate Data on Open Web

Security researchers discovered over 5,000 AI-coded apps from Lovable, Replit, Base44, and Netlify exposing sensitive corporate and personal data without authentication or security measures.

ai llm industry security ai-coding

According to research conducted by cybersecurity firm RedAccess, approximately 40% of the exposed applications contained sensitive information including medical records, financial data, corporate presentations, and detailed customer conversation logs. The applications were discovered through simple search engine queries targeting the domains where these AI coding platforms host user applications.

How AI Coding Platforms Enable Mass Data Exposure

The security vulnerability stems from the default hosting configurations of AI coding platforms. Companies like Lovable, Replit, Base44, and Netlify allow users to host applications on their own domains rather than requiring users to configure separate hosting. This approach simplifies deployment but creates discoverable patterns for automated scanning.

RedAccess researcher Dor Zvi found that many applications had no security controls whatsoever, while others required only trivial authentication such as signing in with any email address. The exposed data included hospital staff assignments with personally identifiable information, corporate advertising strategies, shipping cargo records, and complete chatbot conversation histories with customer contact details.

For European enterprises considering AI coding tools, this research highlights critical due diligence requirements around data governance and security defaults. Teams must evaluate whether these platforms meet GDPR compliance standards and implement appropriate data protection measures.

Platform Responses and Security Responsibility

The AI coding platforms responded defensively to the security findings. Replit CEO Amjad Masad argued that public applications being accessible represents "expected behavior" and that privacy settings can be modified with "a single click." Base44's parent company Wix emphasized that security configuration remains "the creator's responsibility."

Lovable acknowledged taking "reports of exposed data and phishing sites seriously" while noting that secure configuration tools are available but optional. Netlify did not respond to inquiries from Wired regarding the research findings.

These responses reveal a fundamental tension in AI coding platform design: simplifying application creation for non-technical users while maintaining security standards that protect sensitive data. The companies appear to prioritize ease of use over secure-by-default configurations.

Implications for European AI Development Teams

For technical teams evaluating AI coding platforms, this research suggests several operational considerations. First, organizations should establish clear governance frameworks before deploying these tools, particularly in regulated sectors like healthcare and finance where data exposure carries significant compliance risks.

Second, enterprise buyers should evaluate platforms based on their security defaults rather than optional security features. Tools that require explicit security configuration create higher risks when deployed by non-technical users or in rapid prototyping environments.

Third, the discovery method used by RedAccess—simple search engine queries—indicates that exposed applications can be systematically identified and potentially exploited at scale. This differs from traditional application security vulnerabilities that typically require targeted analysis.

Broader Security Architecture Challenges

The mass exposure of AI-coded applications parallels previous cloud security incidents, particularly the widespread misconfiguration of Amazon S3 storage buckets that exposed corporate data from companies including Verizon and World Wrestling Entertainment. Both cases demonstrate how user-friendly tools can create systematic security risks when default configurations prioritize accessibility over protection.

RedAccess estimates that thousands of additional vulnerable applications exist beyond those hosted on platform domains, suggesting the scope extends well beyond the 5,000 applications identified in their initial research. The research methodology—domain-based searches—only captures applications using default hosting configurations.

For European organizations subject to GDPR and other data protection regulations, these findings underscore the importance of conducting security assessments before adopting AI coding tools for applications handling personal or corporate data. The ease of creating web applications with AI tools does not eliminate the need for proper security architecture and data governance frameworks.

Wired's analysis found the security research from RedAccess, though the authenticity of all exposed data could not be independently verified.