EFF Sets Policy for LLM-Assisted Contributions to Open-Source Projects
The Electronic Frontier Foundation (EFF) has introduced a new policy governing contributions to its open-source projects that are assisted by large language models (LLMs). The policy reflects a growing concern about the quality and maintainability of code generated by AI, and aims to ensure the continued high quality of EFF’s software tools.
Balancing Innovation and Quality
The EFF emphasizes its commitment to producing high-quality software, prioritizing substance over simply increasing the volume of code produced. Whereas acknowledging the pervasiveness of LLMs and recognizing that a complete ban would be impractical, the organization is taking steps to mitigate the risks associated with AI-generated code.
Key Requirements for Contributors
Under the new policy, contributors are explicitly required to demonstrate an understanding of the code they submit. Critically, all comments and documentation must be authored by a human. This requirement addresses the potential for LLMs to generate code with subtle bugs or inaccuracies that can be tough to detect, particularly for smaller teams with limited resources.
Addressing the Challenges of LLM-Generated Code
The EFF recognizes that LLMs can produce code that appears human-written but may contain issues like “hallucination, omission, exaggeration, or misrepresentation.” These problems can lead to extensive code reviews that essentially become refactoring exercises for maintainers. The organization also anticipates the potential for a flood of marginally useful or unreviewable AI-generated contributions.
Disclosure is Key
To help maintainers efficiently allocate their time, contributors are asked to disclose when they have used LLM tools in the development process. This transparency allows the EFF to prioritize reviews of well-considered submissions.
Broader Concerns About AI
The EFF’s policy acknowledges broader ethical and societal concerns surrounding LLMs, including issues related to privacy, censorship, and climate impact. The organization points out that these concerns are often rooted in harmful practices within the tech industry. The EFF has previously stated that extending copyright is not a practical solution to the challenges posed by AI-generated content .
A Pragmatic Approach
The EFF’s approach is not about rejecting AI tools outright, but rather about ensuring their responsible use within its open-source projects. The organization advocates for innovation but stresses the importance of understanding how to use these tools safely and effectively.
Read the full EFF policy here.