Gptranslate Multilingual Ai Translation Agent For Wordpress
Monthly
Plaintext third-party AI provider API keys can be extracted from any WordPress site running the GPTranslate - Multilingual AI Translation Agent plugin at version 2.34.6 or earlier, because the plugin emits the stored credential into a public front-end JavaScript asset loaded by enqueue_frontend_scripts. The exposure is unauthenticated and requires no user interaction - an attacker only needs to fetch any public page and apply the inverse transformation that is itself bundled in the plugin's own public JavaScript to recover the key in cleartext (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N, 5.3 Medium as published; our assessment raises the confidentiality component to High given that a raw billing-capable credential is disclosed, yielding 7.5). Keys for OpenAI, DeepL, xAI/Grok, Gemini, Claude, and Google Cloud Translation are all affected, and the default configuration (gpt-3.5-turbo in client mode) is vulnerable; only deepseek-* models and gpt-* models running in server-proxy mode correctly suppress key emission. No public exploit code was identified at time of analysis and the vulnerability is not listed in CISA KEV, though the recovery technique is trivially derivable from the plugin's own shipped assets, so treat exploitation as plausible and rotate credentials.
Plaintext third-party AI provider API keys can be extracted from any WordPress site running the GPTranslate - Multilingual AI Translation Agent plugin at version 2.34.6 or earlier, because the plugin emits the stored credential into a public front-end JavaScript asset loaded by enqueue_frontend_scripts. The exposure is unauthenticated and requires no user interaction - an attacker only needs to fetch any public page and apply the inverse transformation that is itself bundled in the plugin's own public JavaScript to recover the key in cleartext (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:N/A:N, 5.3 Medium as published; our assessment raises the confidentiality component to High given that a raw billing-capable credential is disclosed, yielding 7.5). Keys for OpenAI, DeepL, xAI/Grok, Gemini, Claude, and Google Cloud Translation are all affected, and the default configuration (gpt-3.5-turbo in client mode) is vulnerable; only deepseek-* models and gpt-* models running in server-proxy mode correctly suppress key emission. No public exploit code was identified at time of analysis and the vulnerability is not listed in CISA KEV, though the recovery technique is trivially derivable from the plugin's own shipped assets, so treat exploitation as plausible and rotate credentials.