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
Mitigating Chatbot Service Failures: The Roles of Anthropomorphism, Acknowledgment, and Apology
Juyon Lee, Wujin Chu, Rajat Roy
Journal of Consumer Behaviour · 2026-09-04
This study investigates how human-like (anthropomorphic) chatbot design affects customer responses when chatbot service failures occur, distinguishing between failures in the interaction process versus failures in outcomes. Across five studies with over 1,500 participants, the researchers find that highly anthropomorphic chatbots intensify customer aggression specifically during process failures, because they set stronger expectations for socially competent interaction that are more severely violated. Critically, an apology proves more effective than acknowledging the chatbot's AI nature in reducing aggression after process failures, especially for highly anthropomorphic chatbots. The findings offer actionable guidance for firms designing chatbot recovery strategies and clarify when anthropomorphic design helps versus harms the customer experience.
- Enterprise
- Quality assurance
Research
The Authorship Dilemma: Navigating Copyright Ownership in the Age of Generative AI
Manish Singh, Utkarsh Chadha
International Journal of Law Management & Humanities · 2026-09-04
This article examines how generative AI disrupts copyright law's foundational requirement of human authorship, showing that existing doctrines—originality, work made for hire, joint authorship, and derivative works—produce inconsistent or indeterminate results when applied to machine-generated content. It maps potential claimants (users, developers, platform owners, training data contributors) and finds none satisfies the classical authorship test, making the public domain a plausible default. Comparing approaches across the US, UK, EU, India, China, and Australia, the article concludes that copyright should be reformed to separate authorship, creative contribution, and ownership, proposing a tiered framework calibrated to the degree of human creative control.
- AI policy
Research
Generative AI Updates, Organizational Adaptation, and Enterprise Process Stability
Fang Sun
Journal of Mathematics and Interdisciplinary Applications · 2026-09-04
This paper develops a formal mathematical framework—an impulsive dynamical system—to analyze how periodic generative AI model updates affect enterprise workflow stability. It models the tension between technical improvement and organizational disruption, showing that update frequency and intensity must be jointly optimized rather than chosen independently: frequent small updates reduce peak disruption but impose ongoing coordination costs, while infrequent large updates cause deeper stability losses. Governance investment and organizational adaptation expand the stable operating region, while cognitive overload contracts it. The findings provide actionable guidance for enterprise AI deployment policy, recommending moderate update intervals, moderate update intensity, and relatively high governance investment.
- Enterprise
- Workforce
Research
Certiguard: a fine-tuned LLM for mastering security certification exams
Chun-Ming Lai, Li-Di Chang, Chun-Chieh Chang et al.
International Journal of Information Security · 2026-09-04
CertiGuard is a fine-tuned large language model (LLaMA-2-7B) designed to help learners prepare for cybersecurity certification exams including CISA, CRISC, CEH, and CCNA. Using a two-stage process of continued pre-training on a cybersecurity corpus and supervised fine-tuning with certification-aligned instructions, the system improved average accuracy from 54.1% on a base model to 86.5% using self-consistency Chain-of-Thought reasoning across four exam domains. The framework applies QLoRA-based fine-tuning to reduce computational cost and organizes study into beginner, intermediate, and advanced proficiency tiers. The results demonstrate that domain adaptation and structured prompting can substantially improve LLM performance on professional certification questions, offering a scalable and lightweight alternative to costly traditional training programs.
- Certifications
- Workforce
Research
From HRM practices to AI-enabled innovation: The mediating role of teachers AI readiness
M. Zh. Sabytkhanova, Кайрат Молдашев
Journal of Pedagogical Sociology and Psychology · 2026-09-04
This study of 218 Kazakhstani school teachers finds that strategic human resource management (HRM) practices support teachers' innovative work behaviour partly by building their AI readiness — that is, their capability and preparedness to engage with artificial intelligence tools. Growth mindset independently predicts innovative behaviour but does not amplify the link between AI readiness and innovation. The authors conclude that technology investments in schools must be paired with HRM and professional development practices to be effective.
- Workforce
Research
From Regulatory Authorization to Clinical Accountability: A Control Preventability Framework for Healthcare AI
Tony T. Williams, Amna Jatoi, Sazain Malik et al.
Journal of Global Social Transformation · 2026-09-04
This integrative regulatory and evidence analysis examines the accountability gap surrounding FDA-authorized AI medical devices, finding that among 1,357 cleared devices only 2.5% were linked to registered prospective trials and only 0.2% had evidence evaluating patient-centered outcomes. Key clinical evidence cited includes a randomized trial showing erroneous LLM recommendations reduced physician diagnostic accuracy by 14 percentage points, and a systematic review of 83 studies finding generative AI performed significantly worse than expert physicians in diagnostic tasks. The paper proposes a Control–Preventability Principle that ties accountability to control, foreseeability, and oversight capacity rather than to the location of the final clinical decision, arguing that fragmented frameworks risk placing disproportionate liability on clinicians for harms from systems they do not control. Strengthened prospective validation standards and lifecycle governance across developers, institutions, and clinicians are recommended to close the evidence–accountability gap.
- AI policy
- Certifications
- Quality assurance
Research
Artificial Intelligence Enabled Learning as a Driver of Workforce Readiness in Hospitality and Tourism
Janry Paul A. Santos, Mark Alvin H. Abad
International Review of Management and Marketing · 2026-09-04
This study examines how AI-enabled learning affects workforce readiness among 300 hospitality and tourism students in the Philippines, using structural equation modeling. It finds that AI-enabled learning boosts workforce preparedness both directly and indirectly by increasing students' learning engagement and self-efficacy. The model explains 44% of the variance in workforce readiness, offering evidence that AI-integrated education meaningfully improves employability outcomes in vocational disciplines within an emerging economy context.
- Workforce
Research
Impact of Generative AI on Freelance and Professional Artists: An Empirical Study
Coraline Zoi Visco
International Journal of Social Science Research and Review · 2026-09-04
This empirical study surveyed 80 freelance and professional artists to assess how they perceive the impact of generative AI on their careers and creative work. Findings show that nearly one-third of respondents reported a decline in client opportunities, and most expressed concern that AI models are trained on copyrighted artwork without consent or compensation. Respondents anticipated greater competitive challenges for freelancers in coming years, while believing human creativity and emotional expression remain hard to replace. The study concludes that clear copyright policies and fair compensation practices are needed to protect creative professionals as AI use expands.
- Workforce
- AI policy
Research
Deciding Under Uncertainty: A Systematic Review and Integrated Theories–Contexts–Methods and Antecedents– Decisions–Outcomes Framework of Artificial Intelligence-Driven Decision-Making in Small and Medium-Sized Enterprises
Omar Picone Chiodo, Mazumder Sita, Noptanit Chotisarn et al.
International Review of Management and Marketing · 2026-09-04
This systematic review synthesizes 72 peer-reviewed articles (2018–2025) on how small and medium-sized enterprises (SMEs) use AI-driven decision-making under conditions of environmental uncertainty. Using a TCM–ADO framework, the authors find that AI adoption in SMEs is shaped by technological, organizational, environmental, and leadership factors, and that uncertainty tends to amplify rather than diminish AI's benefits—positioning AI as an adaptive mechanism during turbulence. Outcomes studied include business performance, innovation, sustainability, and resilience, with manufacturing in Asia (particularly China) dominating the research landscape. The review provides practitioners with guidance on treating AI strategically, building complementary capabilities, and maintaining flexible organizational structures.
- Enterprise
Research
EU Influence in Global AI Governance and its Limits
Sina Hoch, Daniel Mügge
Digital Society · 2026-09-04
This paper examines whether the EU's 2024 AI Act will produce a 'Brussels Effect'—the diffusion of EU rules into global AI governance—and finds four structural reasons to be skeptical. The authors argue that the EU faces informational disadvantages, has made pre-emptive concessions that weakened its rules, produces regulations too vague to drive deep cross-border harmonization, and lacks meaningful influence over the leading AI powers, the United States and China. The study combines comparative case analysis with empirical tracking of rule-setting dynamics since the AIA's first legislative draft in 2021. The findings matter for understanding how AI policy actually spreads globally and where the limits of regulatory leadership lie.
- AI policy
Research
AI Agents Can Now Navigate and Complete LMS Tasks: A Call for Pedagogical Innovation
Stavros P. Hadjisolomou, Rita W. El‐Haddad
Intersection A Journal at the Intersection of Assessment and Learning · 2026-09-04
This paper documents that autonomous AI agents can independently log into learning management systems, complete quizzes (in as little as 5 minutes), and fabricate personal reflections on discussion boards—entirely without student involvement. The authors apply Kane's argument-based validity framework to argue that this capability undermines the foundational attribution assumptions behind course grades, program review, and accreditation inferences. The paper reports at least 15 documented agent runs across three platforms and seven tools, and surveys conference attendees to show that most institutions lack settled written guidance on the agentic case. The authors advocate for design-based responses—verified human presence and low-effort faculty changes—over detection-based approaches.
- Certifications
- Quality assurance
- AI policy
Research
Wild, Thick, and Wicked: Situated Evidence on AI-In-Use for Decisions About Deploying AI Systems
Reva Schwartz, G Waters
Social Science Computer Review · 2026-09-04
This paper argues that existing AI evaluation tools—benchmarks, alignment scores, and safety tests—were designed for model development and are ill-suited for assessing how AI systems actually behave once deployed in real-world organizational settings. The authors propose a real-world 'AI-in-use' evaluation framework that treats variability across users, tasks, and contexts as meaningful signal rather than noise, and outlines four design principles for generating deployment-relevant evidence at scale. The framework includes a shared evaluation architecture with a structured observation environment, a metrics hub, and reusable consortium models to connect model capabilities to practitioner and organizational outcomes. This matters for enterprise and policy stakeholders who must make deployment decisions without adequate sociotechnical evidence about value creation, friction, or risk shifts in specific settings.
- Enterprise
- AI policy
Research
Aligning Higher Education Competencies with Industry Evolution: A Study of Futuristic Skills in Tamil Nadu
Vasimalairaja M, K. Maniraja
International Journal For Multidisciplinary Research · 2026-09-04
This mixed-methods study of 1,200 higher education students in Tamil Nadu, India, finds a 'competency paradox': state intervention (the Naan Mudhalvan Scheme) has improved digital literacy, yet critical gaps remain in automation-related skills, problem-solving, and technical reasoning needed for both traditional sectors (automobile, textile, agriculture) and new-age sectors (AI, renewable energy, robotics). Rural and first-generation students face a structural bottleneck largely driven by high certification costs. The findings highlight how socioeconomic and geographic disparities shape workforce readiness in a rapidly evolving labor market.
- Workforce
- Certifications
Research
Artificial intelligence in endometrial and cervical cancer: From ultrasound-based detection to digital pathology and prognostic modeling
Mohamad Tlais, Ibrahim Dhainy, Razan Moghnieh et al.
WArtificial Intelligence in Cancer · 2026-09-04
This minireview synthesizes evidence on AI applications in endometrial and cervical cancer—two diseases accounting for over one million new diagnoses annually—across domains including ultrasound, colposcopy, digital pathology, MRI-based staging, and prognostic modeling. In retrospective datasets, AI has matched or exceeded expert performance on tasks such as ultrasound assessment of deep myometrial invasion and automated cytology triage, and one digital cytology platform has received FDA clearance. However, the review cautions that headline accuracies are frequently inflated by overfitting on small curated datasets, citing a landmark colposcopy model whose near-perfect results were later attributed to uncorrected overfitting by its own senior author. Gaps in external validation, prospective trial evidence, algorithmic equity, and regulatory clarity continue to prevent translation of promising retrospective performance into routine clinical use.
- Quality assurance
- Certifications
Research
The Double-Edged Sword of AI Pair Programmers: A Systematic Literature Review of Security Vulnerabilities in AI-Generated Code and Agentic Development Environments
Mahmoud E. Farfoura, Mohammad A. K. Alia, Ibrahim Mashal et al.
Journal of Sustainable Smart Systems in Education & Environment · 2026-09-04
This systematic literature review synthesizes security evidence from 216 verified records covering AI coding assistants such as GitHub Copilot, ChatGPT-based coding tools, and agentic editors like Cursor. The review documents both beneficial roles—vulnerability discovery, test generation, secure-coding guidance—and recurring weaknesses including injection flaws, weak cryptography, secret exposure, and hallucinated dependencies. Agentic systems introduce additional orchestration-level risks such as indirect prompt injection, tool supply-chain attacks, and permission amplification. The authors propose a unified taxonomy, threat model, and continuous-assurance architecture emphasizing least privilege, sandboxing, provenance tracking, and mandatory human authorization for high-impact actions.
- Quality assurance
- Enterprise
Research
Can artificial intelligence be the answer to green development? evidence from China’s artificial intelligence innovation and development pilot zones
Chuanbo Zhou, Hansha Gu, Zhusan Yang
Frontiers in Environmental Science · 2026-09-04
Using panel data from 282 Chinese cities (2011–2023), this study examines China's National New Generation Artificial Intelligence Innovation and Development Pilot Zones (AIIDPZ) as a quasi-natural experiment. A staggered difference-in-differences approach finds that AIIDPZ designation significantly improves urban green total factor productivity, primarily through industrial structure upgrading and green technological innovation. Effects are stronger in eastern cities, non-resource-based cities, and cities with better digital infrastructure. The findings provide causal evidence that AI-oriented place-based policy can advance green development goals.
- AI policy
- Enterprise
Research
A robust Bayesian and game-theoretic framework for certifying AI-enabled safety-critical systems under structural misspecification
Kazi Md. Tanvir Anzum, Asef Shahriar
Scientific Reports · 2026-09-04
This paper develops a robust Bayesian and game-theoretic framework for certifying AI-enabled safety-critical systems, addressing the challenge that rare failure modes may remain undetected during testing. Using real naval tactical data system (NTDS) failure data, the authors show that standard Bayesian certification can underestimate residual faults by 8.5-fold, while their robust approach—embedding an ambiguity set of up to 54 prior and model configurations in a Stackelberg regulator-developer game—reduces median societal loss by 76–94% depending on threshold settings. The framework identifies three concrete regulatory levers: the likelihood learning-rate floor, the certification threshold, and audit intensity. The findings directly inform how regulators can set sufficiency criteria for AI system testing before deployment approval.
- Certifications
- AI policy
Research
State administration for the use of artificial intelligence in the information security system and improving the defense capability of ukraine
Anastasiia I. Nekriach
Ukrainian Journal of Applied Economics and Technology · 2026-09-04
This article proposes a comprehensive public administration framework for deploying artificial intelligence within Ukraine's information security and defense systems. The author analyzes Ukrainian doctrinal and legislative definitions of AI, public administration, and information security, and benchmarks national regulation against international standards including the EU AI Act, Council of Europe Framework Convention, OECD AI Principles, and UNESCO Recommendations. A seven-component governance model—covering legal, institutional, organizational-managerial, information-analytical, technological, personnel, and international dimensions—is presented, with practical application tied to Ukraine's AI Development Strategy until 2030 and the Defense AI Center 'A1'. The paper argues that integrated implementation of this model will strengthen information security, improve algorithmic transparency and ethics, and enhance defense capability under martial law conditions.
- AI policy
Research
A standards-based systems engineering framework for AI-enabled in-situ monitoring in laser powder bed fusion
Gisuk Hong, Yan Lu, Hyunbo Cho
The International Journal of Advanced Manufacturing Technology · 2026-09-04
This paper presents a systems engineering framework that integrates multiple ISO/IEC and ASTM standards to guide the deployment of AI-based in-situ monitoring in laser powder bed fusion additive manufacturing. The framework formalizes upstream design decisions—such as monitoring-target definition, data partitioning, and latency constraints—before model development, and links them to explicit verification evidence. An evaluation using paired compliant and non-compliant conditions shows that these upstream choices materially affect deployment-relevant capability even when model architecture and capacity are held constant, and that aggregate model metrics can obscure real deployment limitations. The work is relevant to quality assurance and certification because it provides a reusable, standards-traceable structure for converting design decisions into verifiable acceptance criteria for AI monitoring systems.
- Quality assurance
- Certifications
Research
State regulation of artificial intelligence use in ukraine’s defense sector: current state, challenges, and directions for improvement
Т.В. Федоренко, Vira Dabizha
Ukrainian Journal of Applied Economics and Technology · 2026-09-04
This article analyzes the current regulatory landscape for artificial intelligence in Ukraine's defense sector, finding that existing legal frameworks are fragmented, reactive, and lagging behind the rapid deployment of AI technologies in areas such as autonomous systems, intelligence, cyber defense, and decision-support. The authors identify specific gaps in oversight of autonomous systems, human control requirements, risk assessment procedures, and accountability for algorithmically assisted decisions. They propose a comprehensive governance model combining legal, institutional, and organizational mechanisms, including a risk-based approach, clear allocation of authority, interagency coordination, and mandatory human oversight, while balancing defense innovation with legality and human rights protections.
- AI policy
Research
ORGANIZATIONAL INTELLIGENCE IN LUXURY HOTEL HOUSEKEEPING: A STRUCTURAL EQUATION MODELING OF SCRUM AND ARTIFICIAL INTELLIGENCE
Barbara Isabel PONCE PONCE, Mónica Regalado Chamorro, Nancy Guillén
GeoJournal of Tourism and Geosites · 2026-09-04
This study uses structural equation modeling to examine how artificial intelligence, agile Scrum methodology, and organizational intelligence affect operational efficiency, employee satisfaction, and guest service in luxury hotel housekeeping departments across four Latin American cities. Results show significant positive effects of Scrum on operational efficiency (β=0.65) and job satisfaction (β=0.58), AI on operational efficiency (β=0.61), and organizational intelligence as a key mediating variable on both efficiency (β=0.68) and employee satisfaction (β=0.55). The findings provide empirical evidence that integrating AI and agile management practices into hospitality operations can meaningfully improve both operational and human performance outcomes. This matters for the workforce and enterprise domains by demonstrating measurable pathways through which digital transformation and organizational learning improve worker satisfaction and service delivery in luxury hospitality settings.
- Workforce
- Enterprise
Research
Perceived risk of technological displacement and mental health among young workers: evidence from the CSS2023
杨天发, Mao Dai, Shuqi Zhang et al.
Frontiers in Psychology · 2026-09-04
Using nationally representative data from the 2023 Chinese Social Survey, this study finds that young workers (aged 18–35) who perceive a higher risk of being displaced by technology report significantly worse mental health outcomes. The negative association is strongest among workers without higher education, those in the market sector, and urban residents, and is partly explained by elevated unemployment expectations, perceived socioeconomic disadvantage, and pessimistic family-future expectations. The findings demonstrate that the psychological costs of technological change operate through subjective perceptions and social comparison—not just actual job loss—making perceived displacement risk a meaningful psychosocial stressor in its own right.
- Workforce
Research
The Transparency Mandate: The Efficacy of Watermarking and Disclosure Labels in AI-Driven Influencer Marketing
Manish Singh, Utkarsh Chadha
International Journal of Law Management & Humanities · 2026-09-04
This law review article argues that existing advertising disclosure rules are insufficient for AI-driven influencer marketing because they fail to address the synthetic nature of the speaker itself, not just the commercial relationship. The author proposes a 'dual transparency mandate' covering both commercial and ontological disclosure, supported by a layered regulatory structure, distributed liability, standardized mechanisms, and international coordination. The article compares regulatory approaches in the US, EU, and UK, and concludes that a shift from disclosure regulation to transparency governance is necessary to address AI-generated persuasion. The framework includes a registration scheme for commercial AI influencers and an enforcement architecture to operationalize the proposed norms.
- AI policy
Research
CAREER ANXIETY AMONG TOURISM GUIDANCE STUDENTS: EXPLORING FEARS OF AI-DRIVEN TECHNOLOGICAL DISPLACEMENT
Thowayeb H. Hassan, Mostafa A. HASSANIN, Omar M. Ali et al.
GeoJournal of Tourism and Geosites · 2026-09-04
This study examines AI-related career anxiety among 312 tourism guidance students at three Egyptian universities, identifying five dimensions of concern including fear of displacement, economic insecurity, and professional identity threats. Students whose curricula included direct AI application components showed a 0.63 standard deviation reduction in anxiety, and three distinct coping profiles were identified with differing career commitment trajectories. The findings suggest that curriculum reform incorporating hands-on AI instruction can meaningfully reduce anxiety, with broader implications for career counseling and national tourism workforce planning.
- Workforce
- AI policy
Research
Clean Engineering, Unstable Measurement: A Preregistered Reliability Failure of Black-Box LLM Observers on Shared Endpoints
Haoyaun Zhu, Jie Zhang
arXiv · 2026-09-03
This paper audits the reliability of large language model (LLM) 'judges' — models used to score, rank, or filter AI outputs — and finds that they fail basic measurement stability requirements on shared commercial endpoints. Across nearly 53,000 audited requests, same-window repeat rankings achieved only Spearman 0.40 (against a required 0.90), and byte-identical next-day replays reached only 0.78 (against a required 0.99), with the failures attributed to label-to-meaning mapping biases, candidate gaps far below the instrument's noise floor, and non-deterministic outputs from identical inputs. The findings show that a model name on a shared endpoint is not a frozen, reproducible instrument, undermining the validity of preregistered evaluations and leaderboards that treat LLM judges as stable measurement tools. The authors propose a three-level snapshot-identity ladder, eight design rules, and a reporting checklist to help practitioners validate their measurement instruments before fixing evaluation thresholds.
- Quality assurance
- Certifications