Irreversible change
Profound negative long-term changes to social structures, cultural norms, and human relationships that may be difficult or impossible to reverse.
ENTITY
3 - Other
INTENT
3 - Other
TIMING
3 - Other
Risk ID
mit1041
Domain lineage
5. Human-Computer Interaction
5.2 > Loss of human agency and autonomy
Mitigation strategy
1. Implement rigorous, pre-deployment Safety Decision Frameworks (SDFs) with legally binding mechanisms to constrain the development, capability scaling, and autonomous deployment of General-Purpose AI (GPAI) based on potential for systemic societal disruption. Such SDFs must include mandated capability ceilings, independent multi-party deployment authorization, and clear, auditable pause/rollback triggers. 2. Mandate and enforce Human Oversight Mechanisms and Transparency Controls across all critical AI-aided decision systems. This requires the integration of non-circumventable Human-in-the-Loop (HIL) processes for validation, a clear and actionable ability for human users to understand, override, and disengage the system, and deployment of Explainable AI (XAI) to foster trust and retain human agency. 3. Establish a continuous, independent program for Societal Impact Assessment and real-time behavioral monitoring to track long-term, incremental effects on social structures and human relationships. This includes auditing for value misalignment, data drift, and metrics specifically designed to detect the erosion of individual and professional autonomy across all affected domains.