Severe Security Breach: Full Access via a Visible Identifier
Researchers have discovered that simply entering a basic application identifier (app_id) into two of the platform’s APIs was sufficient to register as an authorized user in a private application, despite lacking the necessary permissions.
- api/apps/{app_id}/auth/register – Used for registering a new user with an email and password.
- api/apps/{app_id}/auth/verify-otp – Used for user verification via a one-time password (OTP).
The Problem: The `app_id` is not a hidden value; it is exposed in the URL and the `manifest.json` file.
The Implication: Any external party possessing the `app_id` could perform the registration process, verify their account, and completely bypass authentication mechanisms (including SSO) to gain access to private applications. According to the researchers: “After confirming the email address, we were able to log into the application via SSO, effectively bypassing authentication entirely.” Consequently, theoretical attackers could gain full access to applications that did not belong to them and to all the information stored within them.
Wiz Responds: Rapid Patching – No Evidence of Exploitation
The breach was reported on July 9, 2025, and within 24 hours, Wix released an update that closed the vulnerability. Currently, there is no evidence that the vulnerability was exploited maliciously in the real world. The Base44 breach serves as further proof that AI tools are not inherently “smart” enough on their own, especially when integrated into enterprise software development. Wiz researchers warn that the attack surface of AI tools is expanding, ranging from prompt injection attacks to the leakage of sensitive information.
Recommendations from IPV Security Information Security Experts:
- Reduce Dependency on Visible Identifiers: Avoid exposing identifying values for authorization or user verification. Visible identifiers should be used exclusively for searching public information.
- Secure Sensitive APIs: Every registration, authentication, or application access interface must be protected by strict authorization controls, including Multi-Factor Authentication (MFA).
- Regular Penetration Testing and Code Reviews: Perform periodic security tests to locate misconfigurations before attackers can exploit them.
- Implement Hardened SSO Controls: Do not rely on SSO as the sole line of defense; integrate it with additional controls, such as dynamic and context-based permission management.
- Identity Management and Verification in Development: Embed “Secure by Design” principles into AI tools and development interfaces so that registration and authentication processes are protected from the initial stage.
- Monitoring and Logging: Ensure full documentation of access attempts, including API calls, to enable early detection of suspicious activity.
- Awareness of Organizational AI Risks: Ensure employees understand that AI tools create new attack surfaces that must be managed through meticulous control, strategic planning, and the implementation of smart security measures.
Summary: The Base44 breach was fixed quickly, but it serves as a sharp reminder: while Artificial Intelligence accelerates processes, it also opens new doors for attackers. Only a combination of updates, clear policies, and intelligent controls will allow organizations to leverage the power of AI without paying a heavy price in information security.
**