Implicit Humanization in Everyday LLM Moral Judgments
2026-04-28T07:23:54Z•ac489f8695ee9361a73db18b2158e0c3fea37a5ae8c87f285fc9e079dae65caa
EU AI ActISO/IEC 42001ITSLLMNIST AI RMFPHIauditabilitycompliancecontainerizationegress-filteringgovernancehardeninghealthcarehigh-risk AIisolation-firstisolation-monitoringmodel-hallucinationnetwork-segmentationon-premiseopen-sourcepatient-safetystandards-harmonization
What happened
This feed contains multiple AI policy and systems papers with security- and safety-relevant findings. Most notable: a radiology on-premise LLM deployment that uses an isolation-first, containerized inference stack with strict network segmentation, host-enforced egress filtering, active isolation monitoring and automated hardening tests; the system was approved to process unanonymized PHI but exhibited clinically relevant hallucinations for open-ended generation (repo: https://github.com/ukbonn/ukb-gpt). A separate paper (UGAF-ITS) presents a standards-harmonization framework for Intelligent-–/
Why it matters
A reviewed impact interpretation has not been published for this record.
Evidence and limitations
- Source ID
- arxiv_cs_cy
- Record identifier
- ac489f8695ee9361a73db18b2158e0c3fea37a5ae8c87f285fc9e079dae65caa
- Enrichment time
- 2026-04-28T07:23:54Z
- 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.