Description

A serious Broken Object Level Authorization (BOLA) vulnerability has been identified in Lovable, a widely used AI-driven app development platform. This issue potentially enables unauthorized individuals to access highly sensitive project data from applications created before November 2025. Exposed information may include source code, database credentials, AI-generated chat logs, and even real user data. The flaw allows users on free-tier accounts to interact with backend APIs in a way that bypasses proper authorization checks, effectively granting access to projects they do not own. BOLA vulnerabilities arise when systems fail to verify whether a requester is permitted to access specific resources, making them particularly dangerous and a top concern in API security. Security researchers revealed that a specific API endpoint can return detailed project-related data without enforcing proper access restrictions. The responses reportedly include user identifiers, conversation histories, and internal AI processing details that were never meant for external visibility. Although the issue was disclosed through a bug bounty platform well in advance, it appears that only newer projects have been secured, leaving older ones exposed. This gap creates a prolonged risk window for early users of the platform, whose data may still be accessible through the vulnerable endpoint. The exposure is especially concerning given real-world examples uncovered during analysis. Some affected projects reportedly contained live credentials and personal data tied to organizations and individuals. This raises broader concerns about enterprise risk, as employees from major companies may also be impacted. Users are strongly advised to rotate credentials, secure stored secrets, and treat older project data as potentially compromised. The incident highlights the importance of independent security practices when using rapidly evolving AI development tools.