A Taxonomy of Systemic Risks from General-Purpose AI

AI risks span 13 distinct categories beyond just job displacement: The research identified systemic risks ranging from environmental harm and structural discrimination to governance failures and loss of control, demonstrating that responsible AI oversight requires attention to a much broader landscape of potential harms than commonly discussed.

Knowledge gaps and difficulty recognizing harm are major vulnerability points: The study found that key sources of systemic risk stem from what we don't know about AI systems and our challenges in identifying when harm is occurring, suggesting that transparency and monitoring mechanisms should be central to responsible AI governance.

General-purpose AI's unpredictable development trajectory creates unique challenges: Unlike narrow AI systems, the paper highlights that the unpredictable path of general-purpose AI development makes it difficult to anticipate and prevent systemic risks, underscoring why proactive governance frameworks are needed before problems emerge.

This taxonomy can guide policy priorities: By systematically categorizing 50 contributing sources of systemic risk across 13 areas, the research provides policymakers with a structured tool to identify which risks deserve regulatory attention first, rather than responding reactively to crises.

Systemic risks require society-wide thinking, not just technical fixes: The focus on "large-scale threats that can affect entire societies or economies" emphasizes that responsible AI governance involves coordination across multiple sectors and stakeholders, not just improvements to individual AI systems.