Oct 2026· Zenodo (CERN European Organization for Nuclear Research)
Ethics and Social Impacts of AI
Abstract
Artificial intelligence is rapidly evolving from systems that generate recommendations and content towards autonomous agents capable of planning, reasoning, interacting with external systems, using tools and executing actions with limited human intervention. As AI systems become more agentic, accountability can no longer be understood solely as a legal obligation or post-incident governance exercise. It must also become an operational property embedded into the design, deployment and operation of autonomous AI systems. ACAAI (Accountability by Design for Agentic AI) approaches accountability as a systems engineering property that emerges from the coordinated implementation of organizational and technical controls rather than from any single governance mechanism. Building upon principles from AI governance, cybersecurity, safety engineering, resilience engineering and AI assurance, the framework translates existing governance principles into more than 100 organizational and technical controls spanning the lifecycle of agentic AI systems. ACAAI is structured around six complementary control domains: (1) Organizational Governance, (2) Human Oversight, Consent and Decision Authority, (3) Identity, Authority and Data Governance, (4) Observability, Explainability and Evidence, (5) Runtime Safety and Assurance, and (6) Incident Response and Recovery. Rather than proposing a new regulatory model or prescriptive implementation methodology, ACAAI provides a structured engineering foundation for designing, deploying, operating and retiring accountable autonomous AI systems. The framework also dentifies research and implementation challenges that remain as agentic AI systems become increasingly autonomous, interconnected and collaborative.
GAOKAO-Bench is introduced, an intuitive benchmark that employs questions from the Chinese GAOKAO examination as test samples, including both subjective and objective questions that contribute a robust evaluation benchmark for future large language models and offers valuable insights into the advantages and limitations...
Xiaotian Zhang, Chun-yan Li, Yi Zong et al.· arXiv.org· 216 citations· ⚡17
This work investigates the possibilities of using LLMs in a resume screening setting via a document retrieval framework that simulates job candidate selection and finds that the MTEs are biased, significantly favoring White-associated names in 85% of cases and female-associated names in only 11.1% of cases.
This paper presents a comprehensive overview of the Ultralytics YOLO family, emphasizing architectural evolution, benchmarking, deployment, and emerging directions from YOLOv5 through YOLO27, and examines detection, segmentation, depth, classification, pose, oriented detection, tracking, export, quantization, and deplo...
A novel threat is unveiled in which attackers steer the RAG system's response by injecting malicious passages into its knowledge base, enabling the attacker to steer the response without altering the user input or modifying the RAG weights.
Jiaqi Xue, Meng Zheng, Yebowen Hu et al.· arXiv.org· 109 citations· ⚡8
This work revisits schema linking when using the latest generation of large language models (LLMs) and finds empirically that newer models are adept at utilizing relevant schema elements during generation even in the presence of large numbers of irrelevant ones.
Karime Maamari, Fadhil Abubaker, Daniel Jaroslawicz et al.· arXiv.org· 109 citations· ⚡19
Empirically, PRISM reduces the end-to-end time for data selection and model tuning to just 30% of conventional pipelines, and achieves this efficiency while simultaneously enhancing performance, surpassing models fine-tuned on the full dataset across eight multimodal and three language understanding benchmarks.
Jinhe Bi, Yifan Wang, Danqi Yan et al.· arXiv.org· 73 citations· ⚡4
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduSep 30, 2026
With $2.1 million funding from Google.org, the open-source Public Transit Intelligence Hub will unify public transit monitoring, operations, and passenger communication.