A developer on a Chinese tech forum shared their experience of disabling GitHub Copilot after discovering that code they had written was appearing in suggestions for other users. This phenomenon, while not widely documented, points to a potential privacy flaw in how AI coding assistants learn from user codebases. The developer expressed concern that proprietary or sensitive code could be inadvertently exposed to other developers through the AI's suggestions. While GitHub has stated that Copilot does not use private code from paid users for training, this incident suggests that the model may still surface patterns learned from similar codebases. For overseas developers and engineering teams, this raises important questions about the trustworthiness of AI coding tools, especially in industries with strict intellectual property protections. The post has sparked debate about whether the convenience of AI assistance is worth the potential risk of code leakage, and whether more transparent data handling policies are needed.
A developer's experience with Copilot revealing proprietary code to others raises critical privacy concerns for AI coding tool users.