Exploring Security Risks and Mitigation Strategies in AI Code Helpers

We aim to tackle the urgent concern of security vulnerabilities linked to the utilization of AI code assistants, with a focus on the unintended disclosure of sensitive data such as secret credentials, intellectual property, and other confidential information. With developers increasingly depending on emerging technologies akin to Copilots, Codeium, CodeGeeX, and similar popular AI code assistants, the peril of information exposure emerges as a significant apprehension. This paper delves into the obstacles encountered and potential remedies to alleviate the security risks entwined with these AI code assistants.