Our Origin Story

In 2021, more than a year before the public release of ChatGPT, the EDSAFE AI Alliance was formed to synthesize the initiatives of diverse stakeholders globally and ensure that Artificial Intelligence (AI) innovations in education were aligned and developed with fundamental principles, which ensured safety, accountability, fairness and transparency and ethical development. With a shared vision for clear policy and practice guidance in the rapidly evolving landscape of AI in education, the “SAFE” framework was created. This framework focuses on four key areas of AI development, policy, and implementation practices: Safety, Accountability, Fairness and Transparency, and Effective use of AI. The framework was conceived as a way to foster informed discussion among diverse stakeholders and allow for education to harness AI’s promise while protecting our students and society from its potential peril.

Origins of the SAFE Framework

Artificial intelligence encompasses not just generative AI but a set of artificial intelligence technologies, including surveillance and synthesis AI. As these artificial intelligence tools advance at an incredible pace, policymakers, educators, developers, and industry stakeholders recognized that conventional policy-making frameworks and timelines were insufficient to keep up with technological advancements or the demand for regulatory and enforcement action.

The EDSAFE AI Alliance developed the SAFE framework to address the urgent demand for coherent guidance that can streamline policy development and technology implementation processes, align diverse perspectives, and establish an approach to address both immediate and long-term concerns associated with these technologies. The framework embodies the principle that if AI innovation in education is not conducted safely, with adequate consideration of stakeholder interests, or with inadequate guardrails and guidelines specific to education, the result will be a loss of trust that will stifle or even halt innovation in education altogether.

Global Coherence and Local Adaptability

Recognizing the diversity of educational ecosystems and actors globally, the SAFE Framework incorporates a range of taxonomies and models, drawing from established guidelines such as the GDPR and UNESCO's AI and Education policy guidance. Overall, the SAFE Framework synthesizes 21 global frameworks for trust and technology. By leveraging these resources, the EDSAFE AI Alliance aims to enhance existing systems and empower stakeholders to adapt the framework to their specific contexts, while ensuring global coherence and supporting international best practices.

The Development of a Multi-Stakeholder Solution

By providing a structured approach and shared vocabulary, the framework aims to support policy development. Defining prioritization and clarification steps helps identify critical issues that can be addressed immediately, while also addressing those that require further research and deliberation. Defining areas of responsibility and governance for multiple stakeholders was central to implementing the steps. The issue of safety is established as the framework’s first principle, underscoring its importance while acknowledging the existing regulatory systems in many countries regarding data privacy and security, as well as their role in regulating AI technologies.

The EDSAFE AI Alliance addresses the inherent challenges of moving quickly and purposefully by designing its prioritization and clarification support mechanisms, which aim to guide policymakers through the complexities of AI regulation. Using the SAFE framework, the Alliance has developed a streamlined process and accompanying resources for identifying issues, assessing their urgency, and aligning them with global and national standards, while allowing flexibility to accommodate local requirements and context. This work is continually refined through EDSAFE AI’s policy labs.

Conclusion

The SAFE Framework supports policymakers navigating the fast-moving waters of AI in education. By simplifying complex issues into actionable steps and fostering consensus among stakeholders, the Framework has paved the way for more effective and timely policy implementation, ultimately striving to ensure that AI in education benefits all learners.

The SAFE framework has informed numerous international, national, and local policy discussions, and the EDSAFE AI Alliance continues to develop the framework, support its adoption, and refine its application further through our global alliance work and policy agenda.