Effective Mitigations for Systemic Risks from General-Purpose AI

TL;DR

The paper tackles the challenge of mitigating systemic risks posed by general-purpose AI by combining a literature-derived set of 27 risk-mitigation measures with a survey of 76 domain experts across AI safety, infrastructure, democratic processes, CBRN, and bias. Using a mixed-methods design, the study identifies eight priority measures—most notably third-party pre-deployment audits, safety incident reporting and information sharing, and pre-deployment risk assessments—that experts perceive as both effective across multiple risk domains and technically feasible, with high cross-domain agreement. The findings emphasize the value of external scrutiny, proactive evaluation, and transparency, while warning against reliance on provider self-regulation and suggesting a regulatory emphasis on independent oversight and multi-layered risk strategies. These insights have immediate policy relevance for the EU AI Act and broader governance, offering concrete guidance on which mitigations merit mandating or incentivizing to reduce systemic risks in GP AI systems. The work contributes a novel, domain-specific assessment that blends quantitative consensus with qualitative reasoning, and it calls for further empirical evaluation and practical implementation research to support scalable risk governance.

Abstract

The systemic risks posed by general-purpose AI models are a growing concern, yet the effectiveness of mitigations remains underexplored. Previous research has proposed frameworks for risk mitigation, but has left gaps in our understanding of the perceived effectiveness of measures for mitigating systemic risks. Our study addresses this gap by evaluating how experts perceive different mitigations that aim to reduce the systemic risks of general-purpose AI models. We surveyed 76 experts whose expertise spans AI safety; critical infrastructure; democratic processes; chemical, biological, radiological, and nuclear risks (CBRN); and discrimination and bias. Among 27 mitigations identified through a literature review, we find that a broad range of risk mitigation measures are perceived as effective in reducing various systemic risks and technically feasible by domain experts. In particular, three mitigation measures stand out: safety incident reports and security information sharing, third-party pre-deployment model audits, and pre-deployment risk assessments. These measures show both the highest expert agreement ratings (>60\%) across all four risk areas and are most frequently selected in experts' preferred combinations of measures (>40\%). The surveyed experts highlighted that external scrutiny, proactive evaluation and transparency are key principles for effective mitigation of systemic risks. We provide policy recommendations for implementing the most promising measures, incorporating the qualitative contributions from experts. These insights should inform regulatory frameworks and industry practices for mitigating the systemic risks associated with general-purpose AI.

Figures (7)

• Figure 1: Expert agreement on effectiveness of different risk mitigation measures for general-purpose AI models across two systemic risks. Experts (n=76) were asked to what extent they agreed that the implementation of 27 risk mitigation measures by providers of large general-purpose AI models would effectively reduce four systemic risks from AI. Two are shown here: (1) Disruptions of critical sectors and (2) Negative effects on democratic processes. Experts were also able to indicate that a measure was not yet technically feasible.
• Figure 2: Expert agreement on effectiveness of different risk mitigation measures for general-purpose AI models across four systemic risks. Experts (n=76) were asked to what extent they agreed that the implementation of 27 risk mitigation measures by providers of large general-purpose AI models would effectively reduce four systemic risks. Two are shown here: (3) Chemical, biological, radiological, and nuclear risks (CBRN) and (4) Harmful bias and discrimination. Experts were also able to indicate that a measure was not yet technically feasible.
• Figure 3: Expert agreement on effectiveness of different risk mitigation measures for general-purpose AI models across two systemic risks. Experts (n=76) were asked to what extent they agreed that the implementation of 27 risk mitigation measures by providers of large general-purpose AI models would effectively reduce four systemic risks from AI. Two are shown here: (1) Disruptions of critical sectors and (2) Negative effects on democratic processes. Experts were also able to indicate that a measure was not yet technically feasible.
• Figure 4: Expert agreement on effectiveness of different risk mitigation measures for general-purpose AI models across four systemic risks. Experts (n=76) were asked to what extent they agreed that the implementation of 27 risk mitigation measures by providers of large general-purpose AI models would effectively reduce four systemic risks. Two are shown here: (3) Chemical, biological, radiological, and nuclear risks (CBRN) and (4) Harmful bias and discrimination. Experts were also able to indicate that a measure was not yet technically feasible.
• Figure 5: Most selected measures for reducing systemic risk effectively. Experts were asked what combination of ten measures they believed would be most effective at reducing the systemic risks from general-purpose AI. The Figure shows the percentage of the total experts (n=76) who included each measure in their ten measures (indicated at the end of each bar). The colour of the bar represents the expert group, though note that there were different numbers of experts in each group so that the label of each bar segment does not represent the percentage of the expert group who chose the measure.