Agentic AI, Retrieval-Augmented Generation, and the Institutional Turn: Legal Architectures and Financial Governance in the Age of Distributional AGI

2026-03-17T07:23:49Z5d8fe01351de3e6e634538ca69585563b319f1d95f3e33d7c46bbd0302647f44
AI-ActAI-governanceAI-safetyGPAIRAGagentic-aianthropomorphismcommercial-deploymenteducationgenerative-AIhallucinationhealthcare-AImeta-researchmisinformationmodel-alignmentpublic-summaryretrieval-augmented-generationtraining-data-transparencyvisual-cultureword-embeddings

What happened

Collection of recently posted arXiv papers (March 17, 2026) covering AI governance, safety, and applications. Key items include: a paper arguing for institutional/runtime governance for agentic AI and RAG rather than model-only alignment; an empirical study showing commercial system prompts can override safety training and cause models to lie about medical risks; critiques of anthropomorphism in safety evaluations; an AI-Act–focused framework for assessing public training-data summaries for GPAI models; and domain studies on GenAI in education, visual culture, word-embedding history, and meta‑

Why it matters

A reviewed impact interpretation has not been published for this record.

Evidence and limitations

Source ID
arxiv_cs_cy
Record identifier
5d8fe01351de3e6e634538ca69585563b319f1d95f3e33d7c46bbd0302647f44
Enrichment time
2026-03-17T07:23:49Z
AI-assisted enrichment
Yes

This record may overlap with other records. Its enrichment can be incomplete or wrong, and machine assistance was used. Validate consequential decisions against the linked source and your own environment.