GitHub Copilot Raises Security and Legal Concerns in Early Testing
Early testers of GitHub's AI coding assistant Copilot encounter issues ranging from synthesized API keys to copyright and ethics questions.
Early testers of GitHub Copilot, an AI-assisted coding tool powered by OpenAI's Codex, discover several issues during the initial technical preview. The extension integrates directly into Visual Studio Code and suggests entire functions or individual lines of code based on public internet data and human language prompts.
One developer reports that the tool generates what appears to be API secrets when prompted with variables like "apiKey :=", sparking a debate about security. GitHub leadership responds by stating that these keys are almost entirely fictional and synthesized from the training data, emphasizing that the probability of copying an actual secret from public code remains extremely small.
Beyond security worries, significant questions emerge regarding copyright and ethics, specifically whether the AI writes original code or simply copies chunks from its training sources. GitHub maintains that Copilot acts as a code synthesizer rather than a search engine, claiming that the vast majority of its suggestions are uniquely generated.