The vulnerability allows an unauthenticated remote attacker to modify the AI model configuration and swap the active model during runtime. This bypasses normal authentication controls and can alter the behavior of the inference service, potentially leading to data leakage, incorrect predictions, or delivery of malicious outputs. The weakness stems from the lack of privilege checks on administrative routes and is cataloged as CWE-306.

The affected product is EnergizeLab’s FastAPI AI inference server. No specific version numbers are listed, so any deployment of this product is potentially vulnerable.

The CVSS score of 9.8 indicates a critical level of risk. With no EPSS data available and the vulnerability not currently listed in the CISA KEV catalog, the likelihood of exploitation remains uncertain. However, the nature of the issue—unauthenticated remote administrative access—implies that an attacker with network visibility to the service could easily perform the exploit using standard HTTP requests. The potential impact is high, involving both integrity and confidentiality of the inference results.