Policy for AI-Generated Code

Introduction and Purpose

This policy outlines the guidelines and requirements for the use of Artificial Intelligence (AI) generated code within the Avocado framework project. As AI code generation tools become increasingly prevalent, it is crucial to establish clear standards to ensure transparency, maintain code quality, uphold licensing integrity, and address potential ethical and security concerns. This policy aims to facilitate responsible and effective integration of AI tools while safeguarding the principles of open-source collaboration.

What is AI-Generated Code

Any code, code snippets, configurations, documentation, or other programmatic assets produced, suggested, or significantly modified by an artificial intelligence model, tool, or service. This includes code that is directly used, or heavily adapted from AI suggestions.

Transparency and Disclosure

All contributors are required to be transparent about the use of AI-generated code.

Explicit Attribution: When submitting a Pull Request (PR) or commit that includes AI-generated code, contributors must explicitly state that AI tools were used. This has to be done in the commit message and PR description.

Example Commit Message/PR Description:

 header          <- Limited to 72 characters. No period.
                 <- Blank line
 message         <- Any number of lines, limited to 72 characters per line.
                 <- Blank line
 Assisted-by:    <- artificial intelligence model, tool, or service which
                 has been used and how big part of contribution has
                 been generated (percentage)
 Reference:      <- External references, one per line (issue, trello, ...)
 Signed-off-by:  <- Signature and acknowledgment of licensing terms when
                 contributing to the project (created by git commit -s)

Originality: If AI was used as a brainstorming or refactoring tool, but the final code is substantially original and reviewed by the contributor, a general disclosure in the PR description is sufficient.

Licensing and Copyright

Contributors are responsible for ensuring that AI-generated code complies with the Avocado’s license (GPLv2).

Be Alert to Possible Copyright Issues: Do not assume that AI-generated code is free of licensing obligations or copyright. Treat it with the same diligence as code copied from other sources. To the extent practical, ensure that AI-generated code does not infringe third-party copyrights.
Review for Training Data Leakage: Contributors are expected to enable any code similarity matching or blocking functionality offered by their AI tools if it is available. This helps to identify or prevent potential leakage from proprietary or incompatibly licensed training data. If such instances are identified by the contributor, the contributor must either (a) ensure that the AI-generated code is used in a manner that complies with its license, and ensure that such license is compatible with the license of the project, or (b) the code should be re-written before submission.

Code Quality, Review, and Testing

AI-generated code must adhere to the same quality standards as human-written code.

Human Review Required: All AI-generated code, regardless of its source, must be thoroughly reviewed by a human contributor. This review should be as rigorous, if not more so, than the review of human-written code.

AI contribution best practise

AI as an Assistant: AI tools are encouraged as assistants for boilerplate generation, refactoring suggestions, debugging, or exploring different approaches. They should not replace critical thinking, understanding, and human review.
Understanding is Key: Contributors must fully understand any AI-generated code they submit. Do not submit code you do not comprehend or cannot debug yourself.
Iterative Refinement: Do not blindly accept AI output.

Enforcement and Dispute Resolution

Non-Compliance: Failure to adhere to this policy may result in the rejection of Pull Requests.
Disputes: Any disputes or concerns regarding the use of AI-generated code will be discussed and resolved by the Avocado maintainers.

Disclaimer

This policy is a living document and will be updated as AI technology evolves and best practices emerge. Contributors are encouraged to provide feedback and suggestions to improve this policy over time.