Sep 2026· Zenodo (CERN European Organization for Nuclear Research)
Abstract
Two readings of published instruments, made in September 2026. The first reads four instruments that will oblige deployers to keep recordsagainst a single question: do they require a record to say who authorised aconsequential action, and to be checkable by somebody other than the partythat produced it? The EU AI Act, the ForHumanity certification criteria, theCoSAI Risk Map and the NIST AI RMF. Three of the four specify what a recordmust contain and all three require one to be kept. None requires it to becapable of being shown unaltered by anyone other than its author. NIST isrecorded as not applicable throughout, because a framework that deliberatelyspecifies no controls is not failing to specify one. ISO/IEC 42001 has noverdict because its text is paywalled, which is recorded as a fact about thefield rather than guessed at. The second reads two articles of the EU AI Act against each other. Article 86gives an affected person the right to clear and meaningful explanations of therole of an AI system in a decision and the main elements of the decisiontaken, across seven of the eight areas of Annex III. Article 12 specifies whata log must contain for one of them, remote biometric identification, which isalso the only place the Act requires the record to name the natural personswho verified a result. The duty to explain is therefore seven times wider thanthe duty to record, and the counts show what deployed software can answer: often agent systems read at pinned commits, two can recover the source of astored fact by following a link rather than inferring it, and eight cannot sayafterwards that two stored facts ever disagreed. Neither reading is legal advice. Every verdict names the article, clause,control or criterion it rests on, so that a wrong one is cheap to demonstrate.An earlier version of the first reading said the Act contained no requirementto record a person's identity; that was false, Article 12(3)(d) requires itfor one category, and the correction is recorded on the face of both documentsrather than made quietly. The per-system verdicts these counts come from are deposited separately atdoi:10.5281/zenodo.22290922.
The method, ECCOLA, is presented, which aims at making the high-level AI ethics principles more practical, making it possible for developers to more easily implement them in practice.
Ville Vakkuri, Kai-Kristian Kemell, P. Abrahamsson· EUROMICRO Conference on Soft...· 64 citations· ⚡6
The goal is to not only refine the accuracy of the LLM-based tool but also to underscore its potential in streamlining the software development lifecycle through proactive code improvement and education.
Z. Rasheed, Malik Abdul Sami, Muhammad Waseem et al.· arXiv.org· 62 citations· ⚡3
The use of large language models to automatically improve the user story quality in Austrian Post Group IT agile teams is explored, with a reference model for an Autonomous LLM-based Agent System developed and implemented at the company.
Zheying Zhang, M. Rayhan, Tomas Herda et al.· International Conference on...· 48 citations· ⚡4
This paper introduces a novel multi-AI-agent system designed to fully automate SLRs, and demonstrates how it substantially reduces the time and effort traditionally required for SLRs while maintaining comprehensiveness and precision.
Abdul Malik Sami, Z. Rasheed, Kai-Kristian Kemell et al.· arXiv.org· 44 citations· ⚡2
The proposed LLM-based multi-agent system automates qualitative data analysis process, creating opportunities for researchers and practitioners, and future improvements focus on enhancing multilingual performance and integrating continuous expert feedback.
Z. Rasheed, Muhammad Waseem, Aakash Ahmad et al.· arXiv.org· 41 citations
AI is making software generation faster, but speed does not remove the need for expertise. As more work is delegated to AI, tacit knowledge may become one of the most important human advantages in software engineering. The post Beyond Prompt Engineering: The Role of Tacit Knowledge in Software Engineering appeared first on GPT-Lab.
MIT News · Artificial Intelligence· news.mit.eduSep 16, 2026
The “HardFlow” algorithm could help generative AI models produce high-quality outputs that obey strict requirements when “pretty close” doesn’t cut it.
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.