AI safety is currently defined too narrowly—mostly as a technical problem of making algorithms behave—and this project aims to widen that definition to include economic, social, and environmental harms. The problem is that existing AI safety frameworks, and the new national AI Safety Institutes, focus on risks like algorithmic bias or rogue models, while ignoring the real-world damage AI already causes: job displacement, energy consumption, and the concentration of power in a few tech companies. This project synthesises academic literature, policy documents, and stakeholder perspectives to build what the researchers call an "AI safety net"—a broader framework that captures these overlooked harms. If successful, the project will produce policy recommendations tailored to the UK and Canada, drawing on workshops with industry and civil society groups. It will also create an open-access digital library of annotated resources for policymakers and researchers. The immediate impact is conceptual: shifting how governments and regulators think about AI safety. There is no direct technological output here—this is a knowledge-synthesis and policy-engagement project, not a piece of engineering. But by reframing safety to include labour and environmental consequences, it could influence how future regulations are written and how AI systems are audited.
View original technical description
In response to the increasing international recognition of the harms arising from the deployment of artificial intelligence (AI) systems and the emergent establishment of AI Safety Institutes in multiple countries, this project aims to synthesize current knowledge about AI-related harms and propose an expanded framework of AI Safety. We will critically assess interdisciplinary literature and transnational and national regulatory efforts within an AI safety framework. By bridging academic scholarship, policy documents, and stakeholder perspectives, we aspire to develop what we call an "AI safety net" for the UK and Canada. This project aims to accomplish three goals. 1) Develop the idea of AI safety net by mapping the global landscape of AI safety policies and critically reviewing the existing scholarly literature on the economic, social, and environmental impact of AI production and deployment. 2) Synthesize country-specific approaches and regulatory frameworks for AI safety to develop policy recommendations tailored to the contexts of Canada and the UK. 3) Engage with industry stakeholders and civil society organisations (CSOs) in Canada and the UK to incorporate diverse perspectives for the development of AI safety net policy framework. To achieve our goals of scholarly contributions and policy interventions, our knowledge synthesis will address three interconnected themes: 1) power, dynamics, and competition landscapes shaping AI industry, 2) labour and the workplace in AI production, and 3) environmental implications. We identify these themes based on established literature that has explored the societal, economic, and political consequences of AI, but not within a safety framework, and the emerging scholarly discourse that advocates for a more inclusive examination of AI harms, especially from communities directly impacted by AI. By critically integrating diverse research approaches, our approach aims to transcend the prevailing AI safety paradigm's narrow, predominantly technical orientation. The distinctive value of our approach lies both in developing a more robust framework of AI safety net and in articulating the connections between these three themes to guide future research agendas on AI safety and global governance. This project also distinguishes itself through a comprehensive and synthetic approach to literature review and an inclusive strategy for developing policy recommendations. Our methodology extends beyond traditional academic publications; we will systematically collect and analyze the legal and regulatory documents, government reports, national AI plans, position papers, and research reports and studies by CSOs such as the United Nations, tech industry associations, labour, and environmental organisations. In analyzing this expansive body of non-academic literature, we will proactively reach out to policymaking offices, industry stakeholders, and CSOs to gather their insights and feedback. To further enrich our research, we will host two knowledge mobilisation workshops in Canada and the UK to incorporate stakeholder perspectives into our knowledge synthesis report and policy briefs. Finally, we will create an open-access Zotero digital library that serves multiple purposes. This online resource will feature thematic reading lists, annotated bibliographies, country-specific resources, and regular updates on cutting-edge research and policy initiatives. While at its inception as an academic output for this project, this library serves as a foundational and evolving resource for broader academic and policy communities.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know