The CyRC Vulnerability Advisory has disclosed a critical security flaw in EmailGPT, an AI-powered email writing assistant and Google Chrome extension. This vulnerability, identified as CVE-2024-5184, dubbed prompt injection, permits malicious actors to exploit the service, potentially resulting in the compromise of sensitive data. At the heart of this vulnerability lies the exploitation of the API service, enabling malicious users to inject direct prompts and gain control over the service’s logic.
By coercing the AI service, attackers can force the leakage of standard system prompts or execute unauthorized prompts, opening doors to various forms of exploitation. The implications of this EmailGPT vulnerability extend beyond data breaches. With the ability to submit malicious prompts, individuals can extract sensitive information, initiate spam campaigns, or fabricate misleading email content, contributing to disinformation efforts. Moreover, the vulnerability could lead to denial-of-service attacks and financial losses through repeated requests to the AI provider’s API.
Despite attempts by CyRC to engage with EmailGPT developers through responsible disclosure practices, no response has been received within the stipulated 90-day timeline. Consequently, CyRC advises users to immediately remove the application from their networks, given the severity of the vulnerability, which carries a CVSS score of 6.5 (Medium). As users grapple with this security challenge, staying informed about updates and patches will be crucial in ensuring continued secure service use.
This incident underscores the critical importance of prioritizing security in AI-powered tools. By heeding the recommendations of cybersecurity advisories like CyRC and implementing proactive measures to mitigate risks, users can safeguard their data and uphold the integrity of their digital communication systems in an era of evolving AI threats.
