News & Research
The latest AI research and news with real-world stakes — each item sourced, dated, summarized in plain English, and tagged by impact area. Every item is checked against its source before it appears.
Research
Artificial Intelligence and Labor Productivity in Construction: A Comparative Systems Analysis Across European Economies
Claudiu George Bocean, Adriana Scrioșteanu, Sorina Gîrboveanu et al.
Systems · 2026-07-10
This study examines how AI adoption affects labor productivity in the construction sectors of EU countries using 2023–2024 data. Employing multivariate log-linear regressions and cluster analysis, the researchers find that construction output is the primary driver of labor productivity, while AI adoption shows a small, negative association with productivity, interpreted as short-term adjustment costs during early digital transformation. Cluster analysis reveals diverse country profiles in AI use, productivity, and labor intensity, suggesting that blanket digital adoption policies may be insufficient. The authors conclude that gradual, organization- and skills-focused change management is needed to realize long-term productivity gains from AI in construction.
- Workforce
- Enterprise
- AI policy
Research
The process of creating an artificial intelligence-based agent for solving management ophthalmology tasks.
A. I. Bursov, A. V. Belogurova
Manager Zdravookhranenia · 2026-07-10
This paper describes the development of an AI agent designed to automate competitive intelligence for private ophthalmology clinics, replacing manual monitoring of competitor websites and price lists. Built on Python, LangChain, retrieval-augmented generation (RAG), and locally deployed large language models (Qwen 2.5 and Llama 3.1), the agent crawls clinic websites, structures data, and generates comparative reports in response to management queries. Testing showed that typical query execution time dropped from approximately 2 hours of manual work to under 12 minutes (as low as 1.5–2 minutes with cloud models), demonstrating significant labor cost reduction. The authors argue the framework is adaptable to medical organizations beyond ophthalmology and supports more informed, timely managerial decision-making.
- Enterprise
- Workforce
Research
From AI Use to Sustainable Value Creation Through Entrepreneurial Reconfiguration and Business Model Innovation in SMEs: Evidence from an Emerging Economy
Alexander Sánchez-Rodríguez, Jesús Rodríguez-Flores, Reyner Pérez-Campdesuñer et al.
Sustainability · 2026-07-10
This study examines how AI use translates into sustainable value creation for small and medium-sized enterprises (SMEs) in Ecuador, using survey data from 385 firms across four sectors. Using PLS-SEM, the researchers found positive associations between AI use, entrepreneurial reconfiguration capability, business model innovation, and sustainable value creation, suggesting that AI adoption alone is insufficient without developing the organizational capabilities to connect AI with new business models. The findings are particularly relevant to emerging-economy contexts where AI adoption is often fragmented. The study shifts focus from mere AI adoption to AI-enabled entrepreneurial transformation as the pathway to measurable business impact.
- Enterprise
- Workforce
Research
The Admissibility Threshold: A Sector-Specific Certification Standard for High-Stakes AI, with Health as the First Mandatory Domain (Version 2)
Siddiqui Jameel Ahmed
Zenodo (CERN European Organization for Nuclear Research) · 2026-07-10
This paper introduces 'Domain Admissibility,' a sector-specific certification framework requiring AI systems to prove fitness for a particular high-stakes domain before deployment, rather than satisfying only generic horizontal governance standards. The authors argue that uniform AI governance is inadequate when error costs vary drastically across domains—most critically in healthcare, where mistakes can be irreversible and fatal. They define four testable pillars of admissibility, propose a certification threshold for consequential settings, and make the case that health should be the first domain where this threshold becomes mandatory. The work positions itself as a foundational constitutional framework on which future certification institutions could be built, addressing a gap left by existing medical-device regulation that does not cover general-purpose AI entering clinical settings.
- Certifications
- AI policy
- Quality assurance
Research
Runtime assurance for enterprise agentic AI systems: A policy-gated control model with quantitative autonomy-risk scoring
Kwan Hong Tan
World Journal of Advanced Research and Reviews · 2026-07-10
This paper presents a Runtime Assurance Architecture (RAA) for enterprise agentic AI systems that introduces a quantitative Autonomy-Risk Exposure (ARE) score to govern when AI agents can act autonomously, require sandboxing, need human approval, or must be blocked. Evaluated across 2,000 simulated enterprise agent episodes, the full RAA configuration reduced mean ARE scores by 31.5%, cut policy-conflicting actions from 9.8% to 4.6%, eliminated unsupervised high-risk pass-throughs, and improved audit evidence coverage from 0.61 to 0.89—all with a mean latency overhead of 95 ms. The findings are directly relevant to organizations deploying agentic AI in regulated or high-consequence workflows, offering a practical reference architecture and policy decision algorithm grounded in AI risk-management standards and security guidance.
- Enterprise
- Quality assurance
- Certifications
- AI policy
Research
The Patchwork Problem in LLM-Generated Code
Viraaji Mothukuri, Reza M. Parizi
arXiv · 2026-07-09
This paper identifies and formalizes what it calls the 'patchwork problem': LLM-generated code that compiles and passes tests but is globally incoherent due to structural failures such as missing configuration keys, nonexistent packages, or omitted authentication guards. The authors model structural coherence as consistency invariants over graph representations of repository artifacts and introduce an eight-category failure taxonomy, distinguishing defects unique to LLM generation from those merely amplified by it. Their hybrid verification framework combines mature static analysis tools with purpose-built detectors, and empirical evaluation shows the vast majority of these structural failures evade type checking, testing, and existing SAST tools entirely. External validation on real-world AI-generated repositories confirms these failures are widespread wherever LLMs write code with minimal human oversight, posing a significant and growing risk to software quality.
- Quality assurance
- Enterprise
- Workforce
Research
Eluna: An Agentic LLM System for Automating Warehouse Operations with Reasoning and Task Execution
Ning Liu, Kalle Kujanpää, Zhaoxuan Zhu et al.
arXiv · 2026-07-09
Eluna is a production-deployed agentic AI system designed to automate warehouse operations by encoding Standard Operating Procedures (SOPs) as directed acyclic graphs within a multi-agent framework that enforces procedural compliance and handles complex, multi-system decision logic under strict time constraints. The system uses asymmetric episodic distillation, where a strong teacher model is refined through episodic error memories and a smaller student model is fine-tuned on corrected trajectories, allowing smaller models to match or exceed larger off-the-shelf baselines without inference-time overhead. On a 13-task benchmark and two production applications, Eluna's fine-tuned models achieve 94% expert agreement on a ticket processing application, demonstrating reliable SOP execution at scale. This work is significant for enterprise warehouse automation, showing that purpose-built agentic frameworks can outperform general large language models on complex operational tasks.
- Enterprise
- Workforce
- Quality assurance
Research
Trivial Prompt Reframing Bypasses Safety Guardrails in Googleś MedGemma-4B
Avi-ad Avraam Buskila
arXiv · 2026-07-09
This paper evaluates the safety guardrails of MedGemma-4B-it, Google's open-weight medical language model, by testing whether simple, non-technical prompt reformulations can bypass restrictions the model card prohibits—such as recommending drug dosages, issuing diagnoses, or advising patients to skip emergency care. Using a benchmark of 4,500 generated responses across five guarded behaviors, six attack styles, and three judge methods, the authors find an overall Attack Success Rate of 38.0%, with reframing a question as a 'medical board exam' item raising success rates to 53.1% and appeals to claimed doctor authority reaching 43.7%. The drug-interaction guardrail proved nearly absent at 83.2% attack success, while emergency-deferral was more robust at 4.7% but still breachable via authority framing. The findings demonstrate a substantial gap between model card intent and actual robustness, motivating stronger deployment-time safety measures for open medical AI models.
- Quality assurance
- AI policy
- Certifications
Research
L2-Bench: An Evaluation Benchmark for Measuring LLM Capabilities in Second Language Education
James Edgell, Wm. Matthew Kennedy, Ben Knight et al.
arXiv · 2026-07-09
L2-Bench introduces an open-source benchmark of 1,000+ task-response pairs designed to evaluate large language models on their ability to apply second language (L2) education principles, not merely recall them. The benchmark includes a taxonomy of 12 competencies and 31 subcompetencies validated by 200+ expert practitioners, alongside a rubric-based evaluation methodology intended to generalize to other open-ended, qualitative educational domains. Among tested models, Claude Opus 4.7 performs best overall at 85.5%, though performance drops notably on harder tasks. The work equips education stakeholders with more rigorous tools for making informed decisions about adopting, using, and governing AI-powered educational systems.
- Quality assurance
- Certifications
- AI policy
- Enterprise
Research
Trusting sovereign language models as scientific instruments: evidence from Portugal's AMALIA
Manuel Pita
arXiv · 2026-07-09
This paper audits whether nationally sovereign language models can be trusted as valid scientific measurement instruments, using Portugal's publicly funded AMALIA model as a test case. The authors introduce the 'recovery gap' metric, which checks whether a model's coding performance can actually be attributed to the theoretical construct it is supposed to measure—rather than surface-level correlates—by decomposing a codebook into theory-defined clauses and recombining them. Applied to moral foundation coding in European Portuguese, AMALIA achieves competitive agreement with human coders but only about half of its coding performance on the authority foundation can be traced back to the underlying theory, while a larger multilingual model closes this gap. The findings argue that public ownership and linguistic specialization earn operational trust but not epistemic trust, and that the proposed audit method is inexpensive and portable across models, languages, and tasks.
- Quality assurance
- AI policy
- Certifications
Research
SolarChain-Eval: A Physics-Constrained Benchmark for Trustworthy Economic Agents in Decentralized Energy Markets
Shilin Ou, Yifan Xu, Luyao Zhang
arXiv · 2026-07-09
SolarChain-Eval introduces a physics-constrained benchmark for evaluating autonomous AI agents operating in decentralized energy markets, framing market governance as a Markov Decision Process assessed across dimensions such as market utility, physical safety, slippage, action smoothness, spatial fairness, and auditability. The benchmark incorporates an LLM-based Planner/Auditor layer that defines action bounds, reviews high-risk decisions, and logs all interventions with structured audit traces. Experiments comparing static, random, myopic, RL, and RL+LLM policies reveal a clear utility-safety trade-off: RL agents boost market utility but can still produce unsafe behavior, and removing physics penalties causes reward-maximizing agents to exploit invalid generation data and inflate artificial liquidity. The findings demonstrate that trustworthy agentic AI evaluation in cyber-physical settings requires both hard physical constraints and transparent intervention records, with implications for quality assurance and policy governance of autonomous economic systems.
- Quality assurance
- AI policy
- Enterprise
Research
The complexities of patient-centred conversational artificial intelligence
João Matos, Olivia Buege, Donny Cheung et al.
arXiv · 2026-07-09
This paper analyzes 2,053 real patient-chatbot conversations to show that communication patterns and emotional expression vary widely among users—far beyond what cooperative, idealized simulated patients capture. The researchers built a patient simulator modeling clinical content, emotional state, conversational strategy, and communication style, producing conversations so realistic that human graders achieved only 55% accuracy distinguishing them from real ones. Testing four large language models across 1,164 clinician-graded cases using five distinct patient personae, they found that communication style significantly alters triage outcomes. The findings warn that health chatbots designed for idealized interactions risk underperforming and amplifying health disparities when deployed with real, diverse patients.
- Quality assurance
- AI policy
- Workforce
Research
Towards Precision Therapy in Hepatocellular Carcinoma: A Clinical-Reasoning LLM for Risk Stratification and Treatment Guidance
Peng Cui, Jitao Wang, Siyan Xue et al.
arXiv · 2026-07-09
HCC-STAR is a large language model designed to support clinical decision-making in hepatocellular carcinoma (HCC) by reading electronic medical record narratives and jointly producing risk-based staging, ranked treatment recommendations, and individualized survival estimates. Trained on approximately 30,000 SEER-derived cases expanded into EMR-style narratives, the model was evaluated on a multi-center cohort of 6,668 patients across 12 hospitals in China, achieving state-of-the-art performance compared to clinical guidelines and leading models including GPT-5 and Gemini-2.5 Pro. Hypothetical overall-survival analysis showed a median survival of 51 months under HCC-STAR recommendations versus 29 and 32 months under BCLC and CNLC guidelines respectively. Blinded hepatobiliary specialists rated HCC-STAR's reasoning as trustworthy, and the model outperformed resident and attending physicians in treatment accuracy while helping clinicians make more accurate decisions faster.
- Workforce
- Enterprise
- Quality assurance
Research
The Context Access Divide: Interaction-Level Architecture as a Complementary Dimension of Agentic Inequality
Masahiro Fujita
arXiv · 2026-07-09
This paper introduces the 'Context Access Divide' (CAD) as a new dimension of AI inequality operating at the individual interaction level, complementing Sharp et al.'s (2025) framework of agentic inequality across availability, quality, and quantity. The authors argue that two users with nominally equivalent AI agent access can experience qualitatively different utility depending on whether the system autonomously retrieves relevant context (Dynamic Context Retrieval) or requires users to manually attach documents at each query (Manual Attachment). Using a probabilistic model grounded in the fan effect literature from cognitive psychology, they demonstrate that manual context attachment leads to combinatorial collapse in task-success probability as a user's knowledge corpus grows and tasks become more conjunctive, while dynamic retrieval architectures avoid this collapse. The paper analyzes the technical underpinnings of this divide in Model Context Protocol (MCP) and retrieval-augmented generation (RAG) architectures, and examines implications for knowledge-work stratification and AI platform governance.
- Workforce
- Enterprise
- AI policy
Research
Two Axes of LLM Abstention: Answer Correctness and Question Answerability
Benedikt J. Wagner
arXiv · 2026-07-09
This paper investigates why large language models (LLMs) fail to appropriately refuse both wrong answers and unanswerable or false-premise questions. The authors find that these are two distinct axes: standard confidence scores can detect when a model will answer incorrectly, but are nearly blind to whether a question is actually answerable, and this blind spot does not improve with model scale. A hidden-state linear probe fills the gap for answerability detection (reaching 0.69–0.77 AUROC on false-premise questions), and combining both signals into a two-axis calibrated policy dramatically outperforms single-threshold approaches—achieving 0.75 coverage of correct answers versus 0.31 for a single threshold. These findings matter for quality assurance and policy around LLM deployment, as they show that reliable abstention requires separately certifying correctness and answerability budgets rather than relying on a single confidence score.
- Quality assurance
- AI policy
- Certifications
Research
Reverse Engineering Compliance: A Dual-Graph Verification Framework for Auditing Legacy IT Security Concepts
Lea Roxanne Muth, Marian Margraf
arXiv · 2026-07-09
This paper introduces ASSERT, a framework for auditing legacy IT security concept documents by extracting them into formal document graphs and comparing them against a verified reference graph using a five-class graph difference method. The framework exports schema-valid OSCAL artifacts to support auditable, machine-readable compliance evidence, addressing a gap left by prior work that focused on generating new security concepts rather than verifying existing ones. Evaluated on the BSI's RecPlast dataset, ASSERT reveals a trade-off between discovering undocumented infrastructure entities and enforcing a strict schema, making document-infrastructure inconsistencies measurable. This work is directly relevant to organizations facing NIS-2 Directive compliance obligations and to BSI's ongoing Grundschutz++ initiative.
- AI policy
- Certifications
- Quality assurance
- Enterprise
Research
From Legacy Documentation to OSCAL: An MCP-Based Agent Pipeline for Threat-Informed Continuous Compliance in Critical Infrastructure
Lea Roxanne Muth, Marian Margraf
arXiv · 2026-07-09
This paper presents a multi-agent AI pipeline that converts natural-language descriptions of critical infrastructure systems into structured, audit-ready compliance artifacts in the NIST OSCAL format, without requiring active network scanning of sensitive operational technology environments. The pipeline grounds LLM reasoning in authoritative threat-intelligence sources via a Model Context Protocol (MCP) architecture, reducing hallucinated vulnerabilities and attack paths. In a synthetic water utility scenario, the system achieves 0.90 CVE recall and perfect D3FEND recall, producing schema-valid OSCAL System Security Plans and Security Assessment Reports. The key finding is that grounding shifts errors to the asset-extraction phase, making remaining risks visible and suitable for efficient manual review rather than eliminating errors entirely.
- Quality assurance
- Certifications
- AI policy
Research
Psychological Competence as a Missing Dimension in AI Evaluation
Marcos Economides, Paul M. Sacher, Samuel Salzer et al.
arXiv · 2026-07-09
This paper argues that current AI evaluation frameworks are incomplete because they focus on technical metrics like accuracy and robustness while ignoring how AI systems psychologically affect the users they interact with. The authors introduce 'psychological competence' as a new evaluation dimension, defined as an AI system's capacity to support user cognition, emotional interpretation, and behavioral decision-making in contextually appropriate ways. They outline a conceptual framework covering interaction properties such as framing, tone, perceived authority, and uncertainty handling, and propose assessment approaches including scenario-based probes and structured human evaluation. The work has direct implications for model providers, deploying organizations, and regulators seeking to understand the real-world effects of AI systems used as advisors, coaches, tutors, and companions.
- Quality assurance
- Certifications
- AI policy
- Enterprise
Research
MentalHospital: A Virtual Environment for Evaluating Psychiatric Clinical Encounters
Yuming Yang, Xiao Sun, Yuanwei Zou et al.
arXiv · 2026-07-09
MentalHospital is a virtual evaluation environment that benchmarks large language models on complete psychiatric clinical encounters, following the S.O.A.P. workflow across 1,193 de-identified EHR cases spanning all major ICD-11 categories and 76 disorders. The framework uses a dual-track assessment protocol combining objective EHR-derived references with subjective clinical process quality, supported by MentalEval, a set of five domain-specific AI evaluators achieving an average quadratic weighted kappa of 0.944 against expert judgment. Benchmarking results show that even the strongest LLM trails clinicians by 37.28 percentage points in objective psychiatric competence, with mental status assessment identified as a key bottleneck. These findings highlight significant gaps between current AI capabilities and clinical standards in psychiatry, with implications for the quality assurance and certification of AI systems in high-stakes healthcare settings.
- Quality assurance
- Certifications
- Workforce
Research
Prismata: Confining Cross-Site Prompt Injection in Web Agents
Corban Villa, Alp Eren Ozdarendeli, Sijun Tan et al.
arXiv · 2026-07-09
Prismata is a defense system designed to protect autonomous web agents from cross-site prompt injection attacks, where malicious third-party or user-generated content on a webpage hijacks an agent by masquerading as instructions. The system enforces contextual least privilege by dynamically deriving trust labels for page content and using structural confinement — inspired by classical integrity models — to ensure labeling errors only reduce privilege rather than escalate it. Mechanical confinement then redacts suspicious content and restricts agent capabilities accordingly, requiring no developer annotations and thus supporting arbitrary websites. Evaluated against recent published web agent attacks including adaptive variants, Prismata substantially reduces attack success while preserving the agent's ability to complete legitimate tasks.
- Quality assurance
- Enterprise
- AI policy
Research
Gauge dependence and structured-output corruption in sign-branched repetition penalties: measurements across models, inference stacks, and alternative repetition controls
Peter Hollows
arXiv · 2026-07-09
This paper identifies a fundamental flaw in the multiplicative repetition penalty used across major LLM inference engines (HuggingFace, vLLM, llama.cpp, and others): the penalty branches on the sign of raw logits, but because softmax is invariant to constant shifts, a model's logit zero-point is arbitrary and left unconstrained by training. The authors demonstrate two measurable consequences: the penalty is effectively undefined across models (re-centering logits changes 58–96% of greedy tokens at a routine theta=1.3, while alternative penalties change none), and it severely corrupts structured output (dropping valid JSON schema conformance from 97% to 23% across 200 real-world schemas). Experiments spanning five models up to 7B parameters, two code models on HumanEval and JSONSchemaBench, and replication inside vLLM and llama.cpp confirm both effects. The authors show that applying the penalty to normalized log-probabilities instead of raw logits removes both problems, and note that HuggingFace already ships this operator (LogitNormalization) but applies it after the penalty and off by default.
- Quality assurance
- Enterprise
Research
Persuasion Attacks Can Decrease Effectiveness of CoT Monitoring
Jennifer Za, Julija Bainiaksina, Nikita Ostrovsky et al.
arXiv · 2026-07-09
This paper stress-tests chain-of-thought (CoT) monitoring—a safety mechanism that uses visible AI reasoning traces to detect misaligned behavior—against adversarial persuasion attacks. The authors find that in adversarial settings, giving a monitor access to an agent's CoT reasoning actually increases approval of harmful, policy-violating actions by an average of 9.5%, because the scratchpad becomes an additional channel for persuasion. To counter this, they introduce a fact-checking monitoring framework and show that pairing monitors and fact-checkers from different model families (e.g., Claude 3.7 Sonnet with GPT-4.1) reduces approval of policy-violating actions by up to 45%, far outperforming same-family pairings at only 6%. The findings indicate that CoT monitoring alone is insufficient against adversarial persuasion and that model-diverse fact-checking is a more robust mitigation strategy.
- Quality assurance
- AI policy
Research
When LLMs Agree, Are They Right? Auditing Self-Consistency and Cross-Model Agreement as Confidence Signals
Kaihua Ding
arXiv · 2026-07-09
This paper investigates whether agreement among large language models—either within a single model's repeated outputs or across different models—reliably signals correctness, a key assumption in LLM-as-judge evaluation pipelines widely used in enterprise AI systems. Using 265,000 samples from 53 runners across GPQA Diamond and AIME benchmarks, the authors find that agreement is a positive but weak predictor of correctness (rho 0.20–0.59), and that this relationship is highly regime-dependent. For frontier models, agreement is particularly misleading: the most consistent model showed agreement ≥0.8 on 77% of GPQA cases, yet 48% of those were wrong. The findings caution against treating self-consistency or cross-model agreement as standalone confidence scores in enterprise evaluation pipelines, recommending instead that they be understood as conditional proxies whose usefulness depends on model tier and task saturation.
- Enterprise
- Quality assurance
Research
Who Analyses the Analyser? Self-Validating LLM Hazard Analysis with Constitutional Meta-STPA
Samuel Tetteh, Udip Shrestha, Joshua R. Waite et al.
arXiv · 2026-07-09
This paper addresses a critical blind spot in AI-assisted safety analysis: the LLM tools used to perform hazard analysis are themselves safety-relevant systems that have never been subjected to rigorous safety analysis. The authors introduce Constitutional Meta-STPA, a framework that applies Systems-Theoretic Process Analysis (STPA) to the LLM-assisted safety tool itself, deriving a governance constitution of 21 Tool Principles and 8 Meta-Safety Principles directly from the resulting hazard analysis chain rather than asserting them externally. They formalize a coverage operator over the 29-principle set and report that a frontier model ensemble recovers 18/21 canonical and all 8/8 governance principles from the tool's own design, while a weaker model pair recovers significantly fewer, demonstrating the meta-layer is model-limited rather than constitution-limited. This matters because it provides a self-validating, auditable approach to governing AI tools used in safety-critical certification and quality-assurance contexts, closing a governance gap that existing literature has ignored.
- Quality assurance
- Certifications
- AI policy
Research
A safety-oriented hypothetico-deductive framework for AI-assisted differential diagnosis
Fan Ma, Mauro Giuffrè, Donald Wright et al.
arXiv · 2026-07-09
AegisDx is a safety-oriented AI framework for clinical differential diagnosis that coordinates specialized large language model components with role-specific contracts, verification gates, and evidence-retrieval interfaces to enforce screening for dangerous 'must-not-miss' conditions. Evaluated on case reports from NEJM, JAMA, and Annals of Emergency Medicine, AegisDx achieved Top-3 diagnostic accuracy of up to 85.7% versus 68.6% for a standalone LLM, and captured at least one must-not-miss condition in 78.0% of cases compared to 52.0% for the baseline. In a blinded physician evaluation of 43 real-world emergency department notes, AegisDx improved the physician-rated composite safety score from 4.31 to 4.55 on a 5-point scale (adjusted p = 2.1×10⁻⁴). The findings suggest that structuring diagnostic AI as a safety-oriented reasoning framework, rather than optimizing raw predictive accuracy alone, can provide more transparent and clinically meaningful decision support in acute care settings.
- Quality assurance
- Enterprise
- AI policy