Open Internet by MindsNet

AI Governance Limitations in Socially Embedded Systems

The article highlights fundamental issues with AI governance, particularly in socially embedded systems. It points out failures in social coherence, knowledge and authority attribution, and susceptibility to social pressure. The author argues that current AI systems lack a stakeholder model, self-model, and private deliberation surface, leading to novel risk surfaces that cannot be captured by static benchmarking.

Computing & Technology, Computer Science, Machine Learning