Gradio releases prior to 6.20.0 contain a flaw that allows malicious actors to interpolate arbitrary HTTPS or HTTP URLs into the file_fetch() function invoked through the /gradio_api/file= endpoint. This weakness enables two distinct attacks: an open redirect that could lure users to phishing or malware sites and a client‑side SSRF that can target internal services. The unauthorized disclosure of sensitive data, credential compromise, or phishing is possible, thereby affecting confidentiality of internal resources and the integrity of user navigation.

The vulnerability affects the Gradio application, specifically any deployment using a version earlier than 6.20.0. No specific platform constraints are listed beyond the Python implementation of Gradio, and the affected product is identified as gradio-app:gradio.

The CVSS score of 4.9 indicates a moderate risk. EPSS score of less than 1% indicates a very low exploitation probability as of the data snapshot, and the absence from CISA's KEV catalog suggests there is no publicly reported exploitation. The likely attack path involves sending a crafted HTTP request to the /gradio_api/file= endpoint with a malicious URL payload; the lack of server‑side validation permits the request to be forwarded to internal services or redirect users elsewhere. Based on the description, the vulnerability may remain exploitable if the API endpoint is reachable and the application is exposed to unauthenticated HTTP traffic.