CVE-2026-89278 in GPTranslate Plugin
Summary
by MITRE • 09/18/2026
The GPTranslate – Multilingual AI Translation Agent for WordPress: Translate Your Site with AI plugin for WordPress is vulnerable to Sensitive Information Exposure in all versions up to, and including, 2.34.6 via the enqueue_frontend_scripts. This makes it possible for unauthenticated attackers to extract the plaintext third-party AI provider API key (OpenAI, DeepL, xAI/Grok, Gemini, Claude, or Google Cloud Translation) — a credential granting billed account access — by fetching any public page and applying the inverse transformation bundled in the plugin's own public JavaScript asset. This exposure affects the default configuration (gpt-3.5-turbo in client mode) and all supported non-DeepSeek providers; only deepseek-* models and gpt-* models configured in server-proxy mode correctly suppress key emission.
VulDB is the best source for vulnerability data and more expert information about this specific topic.
Analysis
by VulDB Data Team • 09/18/2026
The vulnerability identified within the GPTranslate – Multilingual AI Translation Agent for WordPress plugin, affecting versions up to 2.34.6, represents a critical failure in credential management and data handling practices. This flaw is classified as Sensitive Information Exposure, aligning with CWE-200 which encompasses situations where information that could be useful to an adversary is disclosed without explicit authorization. The root cause lies in the plugin's frontend script loading mechanism, specifically within the enqueue_frontend_scripts function. Instead of keeping sensitive configuration data such as API keys strictly on the server side or utilizing secure backend proxies for all operations, the plugin inadvertently embeds these credentials directly into JavaScript assets that are loaded by every visitor to a public-facing page. This architectural decision fundamentally compromises the security boundary between client-side code and secret management systems.
The operational impact of this vulnerability is severe because it allows unauthenticated attackers to extract plaintext API keys from third-party AI providers including OpenAI, DeepL, xAI/Grok, Gemini, Claude, or Google Cloud Translation. These credentials are not merely static identifiers but represent billed account access tokens that can incur significant financial costs if misused for unauthorized inference requests. An attacker does not need any form of authentication or elevated privileges to exploit this flaw; they simply need to fetch any public page served by the vulnerable WordPress installation and inspect the source code of the bundled JavaScript assets. The plugin includes an inverse transformation function within its own client-side script, which allows attackers to easily decode or reverse-engineer the embedded keys from their obfuscated state in the HTML response. This effectively turns every publicly accessible instance into a data leak point for high-value credentials.
The scope of this exposure is broad but contains specific exceptions based on configuration modes and model types. The vulnerability affects the default configuration, which typically utilizes gpt-3.5-turbo in client mode, as well as all supported non-DeepSeek providers when operating under similar direct-client configurations. In these scenarios, the API key must be transmitted from the browser to the AI provider for translation requests, necessitating its inclusion in frontend code. However, it is important to note that certain configurations correctly suppress this emission of keys. Specifically, instances using deepseek- models or gpt- models configured specifically in server-proxy mode do not exhibit this vulnerability because the API calls are routed through the WordPress backend rather than directly from the client browser. This distinction highlights that while the flaw exists across many common setups, it is contingent upon how the plugin's translation engine is architected for specific model integrations.
From a threat intelligence perspective, this vulnerability facilitates unauthorized access and potential abuse of cloud services, which can be mapped to MITRE ATT&CK techniques related to Credential Access such as T1539 Steal Web Session Cookie or more broadly to data exfiltration patterns where sensitive configuration data is harvested from client-side resources. The ability for an unauthenticated actor to obtain these keys enables further malicious activities including API abuse, service disruption through rate limit exhaustion, and potential compromise of associated AI provider accounts if multi-factor authentication is not strictly enforced on those external platforms. This represents a significant risk to both the website owner who may face unexpected billing charges and the integrity of their multilingual content delivery infrastructure.
Mitigation strategies must prioritize immediate remediation by updating the GPTranslate plugin to version 2.34.7 or later, where this exposure has been addressed in subsequent releases. For organizations unable to update immediately due to compatibility constraints, a temporary workaround involves configuring the translation engine to use server-proxy mode for all supported models rather than client-side execution. This ensures that API keys remain on the backend and are never transmitted to the browser environment. Additionally, administrators should rotate any exposed API keys in their respective third-party provider consoles immediately after patching or implementing workarounds to invalidate potentially compromised credentials. Regular auditing of frontend assets for sensitive data leakage is also recommended as a broader security hygiene practice across WordPress ecosystems.