The Generative AI Commons is dedicated to fostering a collaborative ecosystem for open source projects and open data sets that adhere to the principles of responsible AI. To ensure the ethical development and use of generative AI technologies, we have established the following policy for projects and data sets hosted and promoted by our organization.
Fairness
All open source projects and data sets must prioritize fairness and aim to mitigate biases. Projects should be designed to treat all users equitably and data sets should be representative of diverse perspectives, avoiding the perpetuation or exacerbation of existing biases.
Guidelines:
- Conduct thorough bias assessments during project development and data set curation.
- Implement strategies to reduce and mitigate biases in AI systems and data sets.
- Ensure that AI systems provide fair outcomes for all users, irrespective of their race, gender, age, or other characteristics.
Transparency
Transparency is a crucial aspect of responsible AI. Open source projects should provide clear explanations and comprehensive documentation, while open data sets should be accompanied by relevant metadata and context.
Guidelines:
- Document the purpose, scope, and technical specifications of AI systems and data sets.
- Provide explanations of the algorithms, data processing methods, and decision-making processes employed in AI systems.
- Ensure that users can easily access and understand information about AI systems and data sets.
Accountability
Developers and data set curators must take responsibility for the potential consequences of their AI systems and data sets, designing them with mechanisms for monitoring, evaluation, and redress.
Guidelines:
- Assess the potential social, ethical, and environmental implications of projects and data sets during development and curation.
- Establish processes for monitoring and evaluating the performance and impact of AI systems and data sets.
- Implement mechanisms for redress and address any unintended consequences or negative outcomes.
Privacy and Security
AI projects and data sets must prioritize data privacy, confidentiality, and security, adhering to relevant data protection laws and best practices.
Guidelines:
- Ensure compliance with data protection laws, such as the GDPR, and follow industry best practices for data security.
- Anonymize personal data in data sets and implement data encryption, where necessary.
- Regularly review and update security measures to protect against emerging threats and vulnerabilities.
Collaborative and Open Development
Projects and data sets should encourage open collaboration, knowledge sharing, and the integration of diverse perspectives and expertise.
Guidelines:
- Develop projects and curate data sets using open licenses that encourage collaboration, reuse, and adaptation.
- Foster a community-driven development process, inviting contributions and feedback from diverse stakeholders.
- Promote knowledge sharing by making project documentation, source code, and data sets accessible to the AI community.
