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
Gender-based discrepancies in the algorithmic delivery of political ads on social media
Dominik Bär, Francesco Corso, Gianmarco De Francisci Morales et al.
arXiv · 2026-06-09
This study examines gender-based discrimination in the algorithmic delivery of political ads during the 2024 European Parliament elections, drawing on a large-scale dataset of over 110,000 ads from 453 political parties and 968 candidates that generated over 7 billion impressions across 25 EU countries. The authors find that men were significantly more likely than women to be shown ads from populist and far-right parties—on average a 6 percentage point higher male share—even after controlling for ad content, platform-level competition, and targeting strategies. These algorithmic imbalances restrict parties' ability to reach diverse audiences and prevent voters from engaging equally with the full spectrum of political viewpoints. The findings call on platforms and policymakers to audit algorithmic ad delivery and implement safeguards to protect fairness and democratic processes.
- AI policy
Research
The Arbiter Agent: Continually Monitoring Multi-Agent Conversations to Detect Emergent Misalignment
Filippo Tonini, Federico Torrielli, Anton Danholt Lautrup et al.
arXiv · 2026-06-09
This paper introduces the Arbiter, a monitoring agent designed to watch multi-agent AI conversations in real time and identify which participants are behaving in misaligned ways. Operating under a limited inspection budget, the Arbiter can wait, question participants, examine internal information like system prompts or reasoning traces, or log concerning behavior, ultimately producing a report on the likely source of misalignment. Evaluated across five conversation conditions—including risky financial advice scenarios, evaluation-aware agents, and colluding agents—the Arbiter reliably detects misaligned agents well before conversations end, with active inspection tools improving both detection accuracy and speed. The findings suggest that continual, budget-aware monitoring is effective and that auditing multi-agent systems may require treating the auditor as an active participant rather than a passive observer.
- Quality assurance
- AI policy
Research
The Agentic Web Requires New Normative Infrastructure
Cameron Pattison, Matthew Boulos, Noam Kolt et al.
arXiv · 2026-06-09
This paper argues that AI agents acting on behalf of users on the internet (the 'agentic web') are now technically feasible but face legal and normative barriers, as existing laws, terms of service, and platform practices often block or degrade agent access without distinguishing between malicious bots and legitimately authorized user agents. The authors contend that realizing the social benefits of such agents requires not just technical protocols but a new normative infrastructure—broadly accepted laws, norms, and practices governing agentic access to online platforms. The paper aims to initiate a societal conversation, identify guiding normative principles, and advocate for policies that enable users' delegated agents to act online with minimal unreasonable restrictions.
- AI policy
Research
Hidden Consensus:Preference-Validity Compression in Human Feedback
Dorcas Chia Ern Chua, Karen Myn Hui Lee, Jia Yue Tan et al.
arXiv · 2026-06-09
This paper identifies a flaw in standard Reinforcement Learning from Human Feedback (RLHF) pipelines, which the authors call 'Preference-Validity Compression': the collapse of multiple culturally or normatively valid response options into a single optimization target. Using Malaysia as a test case, the researchers analyzed 321 preference events from 20 participants across 107 annotated prompts, finding that 79% of prompts contain more than one majority-supported response that single-winner aggregation would discard. The study argues this is a measurement-validity problem rather than annotation noise, and proposes that future alignment methods satisfy 'Validity-Preserving Consistency' — remaining stable across plural-valid interpretive frames. The findings matter for AI policy and quality assurance because they reveal that standard RLHF may systematically mis-measure alignment in structurally plural, multicultural societies.
- AI policy
- Quality assurance
Research
Stop Early, Spend Less: Hidden-State Probes as a Practical Recipe for Streaming Moderation of LLM Outputs
Huizhen Shu, Xuying Li, Piao Xue
arXiv · 2026-06-09
This paper proposes lightweight token-level probes trained on LLM hidden states to perform real-time safety moderation during text generation, rather than after it completes. By reusing internal activations from the generator model, the probes require no additional forward pass and enable sub-millisecond per-token safety checks, achieving orders of magnitude lower compute overhead compared to post-hoc or streaming guard models. A probe applied to a single mid-layer can recover most decisions of a strong guard model, allowing unsafe outputs to be halted or modified before generation finishes. The work also provides a practical deployment recipe covering layer selection, aggregation strategy, probing frequency, and triggering thresholds, making it directly applicable to user-facing LLM systems.
- Enterprise
- Quality assurance
Research
Trace2Policy: From Expert Behavior Traces to Self-Evolving Decision Agents
Junli Zha, Jinbo Wang, Chao Zhou et al.
arXiv · 2026-06-09
Trace2Policy introduces EISR (Error-driven Iterative Skill Refinement), a system that recovers tacit decision rules from expert behavior traces in compliance-sensitive domains like auditing and contract review, then iteratively improves those rules through error clustering and targeted patching. Deployed over 22 days at a major logistics carrier across 3,349 audit cases, the compiled Python pipeline achieves 79.6% accuracy after eight refinement rounds, outperforming the pure-LLM baseline it replaced (72.7%), with zero LLM calls at inference. The paper's key finding is that rule quality—not model capability—is the dominant performance lever for skewed-base-rate compliance tasks, and that compiled execution runs 9.8 percentage points higher than prompting the same rules through an LLM. An automated variant (Auto-EISR) replicates the refinement cycle at $5–$10 per cycle versus approximately 70 expert-hours, and transfers to public benchmarks including LegalBench and BPIC 2012.
- Enterprise
- Quality assurance
Research
Agentomics: Economic Foundations for the Valuation, Attribution, and Pricing of AI Agents in Human-AI Workflows
Quanyan Zhu
arXiv · 2026-06-09
This paper introduces 'Agentomics,' a formal economic framework for valuing, attributing, and pricing AI agents within human-AI workflows. Rather than measuring isolated technical performance, it models workflows as configurations of heterogeneous agents whose collective output determines gross value, deployment cost, reliability, and failure risk. It applies the Shapley value from cooperative game theory to fairly attribute economic surplus among AI agents, and derives a 'Shapley pricing equilibrium' as a normative benchmark for whether agent prices reflect their marginal contribution. A security-operations case study illustrates the framework's application to hybrid human-AI workflows involving productivity gains, deployment costs, and reliability trade-offs.
- Enterprise
- Workforce
Research
Position: Preventing AI-Generated CSAM Necessitates New Approaches to AI Safety
Neil Kale, Rebecca Portnoff, Pratiksha Thaker et al.
arXiv · 2026-06-09
This position paper argues that protecting children from AI-facilitated sexual abuse requires fundamentally new AI safety approaches, because existing techniques—such as dataset auditing, red teaming, and fine-tuning prevention—assume levels of data accessibility and transparency that are ethically and legally incompatible with child sexual abuse material (CSAM). The authors identify 15 open problems spanning the full AI development lifecycle, from dataset curation and model design through deployment and long-term maintenance. They offer targeted recommendations for researchers, developers, and policymakers to translate responsible AI principles into concrete safeguards, framing child protection as a central, safety-critical dimension of AI research.
- AI policy
- Quality assurance
Research
Atomic Intent Reasoning: Bringing LLM Semantics to Industrial Cross-Domain Recommendations
Zhuohang Jiang, Yuxin Chen, Shijie Wang et al.
arXiv · 2026-06-09
This paper presents AIR (Atomic Intent Reasoning), a cross-domain recommendation framework that uses large language models (LLMs) to bridge semantic gaps between content and e-commerce platforms. By shifting LLM inference to an offline phase and dynamically composing user intent representations during online operations, AIR achieves approximately 400x inference acceleration while preserving semantic quality. Deployed in Kuaishou E-commerce, the system delivered measurable business gains including a +3.446% increase in GMV in large-scale A/B testing, validating its practical value for industrial-scale recommendation systems.
- Enterprise
Research
Catching One in Five: LLM-as-Judge Blind Spots in Production Multi-Turn Transaction Agents
Sawyer Zhang, Alexander Wang, Sophie Lei
arXiv · 2026-06-09
This paper evaluates the reliability of LLM-as-judge systems for quality assurance in a deployed multi-turn food-and-beverage ordering agent, using exhaustive human transcript review as ground truth. The authors find that the built-in LLM judge catches well under a quarter of genuine quality problems — as few as 2 of 9 confirmed defect patterns (22%) in one batch, and zero flagged failures in a batch where humans confirmed 23 distinct defects. The failure is structural: the judge's rubric covers only coarse axes like intent and brand-voice, leaving state-tracking, guardrail, and recovery defects entirely undetected, and a routing-and-wiring flaw means even defects the judge's raw notes describe are never escalated to operational alerts. The authors conclude that automated LLM judging functions as a regression floor at best and cannot substitute for human review in production multi-turn agents, with statistical corrections implying a 3–6x undercount of true defect rates where any signal exists at all.
- Quality assurance
- Enterprise
Research
MIRAGE: A Polarity-Flipping Encoding Subspace in LLM Agents
Pratibha Revankar, Kargi Chauhan, Jihye Kim et al.
arXiv · 2026-06-09
This paper identifies a shared low-dimensional 'encoding subspace' in the internal representations (residual stream) of large language model agents that activates when the model is covertly encoding sensitive data using schemes like Base64, ROT13, or acrostic ciphers. A logistic-regression probe trained on eight encoding families generalizes to a held-out ninth with AUC 0.975–1.000, and a two-channel real-time monitor called MIRAGE reaches AUC 0.918 on 126 agentic data-exfiltration scenarios, far outperforming output-only detection (AUC 0.518). The work also finds that the same internal direction flips polarity at the planning stage to distinguish whether the model will execute encoding itself or delegate it to a tool call, enabling detection before the encoded output even exists. These findings matter for AI safety and quality assurance because they demonstrate that monitoring model-internal geometry can catch covert exfiltration attempts that evade surface-level output filters, though reliability varies substantially by model architecture.
- Quality assurance
- AI policy
Research
Supervised Fine-tuning with Synthetic Rationale Data Hurts Real-World Disease Prediction
Buxin Su, Bingxuan Li, Cheng Qian et al.
arXiv · 2026-06-09
This paper tests whether fine-tuning language models on synthetic rationale data—explanations of why a prediction is correct—improves clinical disease prediction, specifically five-year Alzheimer's disease and related dementias (ADRD) forecasting from longitudinal health records. Across 504 controlled configurations, the authors find that rationale-based supervised fine-tuning consistently hurts prediction performance compared to label-only fine-tuning, and this degradation holds across model families and data scales. Notably, the failure is not due to low-quality rationales: human experts confirmed the rationales were medically accurate, and the same rationales improved performance when used at inference time rather than as training targets. The authors attribute the problem to a structural conflict between narrative plausibility and discriminative optimization, cautioning against the widespread assumption that rationale-based supervision benefits high-stakes clinical prediction tasks.
- Quality assurance
- AI policy
Research
Artificial Intelligence Governance in Indonesian Education: Regulatory Analysis and the Strengthening of Academic Integrity in the Era of Generative AI
Rizki Auliadi, Mikraj Mikraj
Jurnal Ilmiah Guru Madrasah. · 2026-06-09
This study examines how Indonesia currently regulates AI in education and finds that existing rules are fragmented and sector-specific, lacking a unified legal framework for educational settings. Drawing on normative legal analysis and comparisons with international regulatory practices, the authors identify transparency, accountability, data protection, fairness, and human oversight as core governance principles. The paper proposes a five-pillar model to strengthen academic integrity—covering institutional policy, AI literacy, transparency, adaptive assessment, and monitoring—intended to guide the development of ethical and responsible education policy amid rapid AI advancement. The findings are relevant to policymakers and educators seeking to address risks such as AI-assisted plagiarism, information fabrication, and declining critical thinking skills.
- AI policy
- Certifications
- Quality assurance
Research
Impact of artificial intelligence adoption on corporate green innovation under environmental regulation and subsidies
Chang Dou, Chang Liu (35901), Jiarui Li et al.
Discover Sustainability · 2026-06-09
This study analyzes panel data from Chinese A-share manufacturing firms (2016–2022) to assess how AI adoption affects corporate green innovation, measured through authorized green patents. Results show that AI adoption significantly promotes green innovation, with stronger effects in eastern and central regions, less-polluting firms, state-owned enterprises, and high-technology industries. Mechanism analysis finds that environmental subsidies amplify the positive effect while environmental regulations weaken it, suggesting the need for better policy coordination between AI incentives and green innovation frameworks.
- Enterprise
- AI policy
Research
Artificial intelligence applications and researchers’ wages: from the perspective of R&D resources optimization
Ying Wu, Xi Wu
Science and Public Policy · 2026-06-09
Using data from Chinese listed firms from 2017 to 2023, this paper finds that artificial intelligence applications have a positive impact on researchers' wages. The mechanism appears to work through reductions in non-labor R&D costs and improvements in the human capital structure of R&D teams. The positive effect is weaker for private firms and high-tech firms, while regional high-skilled labor supply does not significantly moderate the relationship. The findings contribute novel evidence on how AI shapes compensation and employment in high-skilled labor markets.
- Workforce
- Enterprise
Research
The nuclear analogy in AI governance research
Sophia Hatz
Edward Elgar Publishing eBooks · 2026-06-09
This chapter reviews 43 scholarly works that use nuclear weapons as an analogy for AI governance, identifying four problem areas where researchers apply nuclear precedents: early development of transformative technologies, international security risks, international institutions and agreements, and domestic safety regulation. The authors argue that even where technological domains differ substantially, nuclear analogies can still inform policy by providing conceptual frameworks for strategic dynamics, cautionary lessons about failed governance approaches, and inspiration for radical policy proposals. Because policymakers already invoke the nuclear analogy, the authors conclude that continued critical engagement with these historical precedents is essential for shaping effective global AI governance.
- AI policy
Research
Multilateralism in the global governance of artificial intelligence
Michał Natorski
Edward Elgar Publishing eBooks · 2026-06-09
This chapter analyzes how international multilateral institutions and frameworks are responding to AI as a general-purpose technology, identifying key principles—epochal change, determinism, and dialectical understanding—that underpin AI governance discussions. It finds that AI issues have been integrated into existing cooperation frameworks while new AI-specific frameworks have also been created. Despite multi-stakeholder appearances, states remain the dominant decision-makers in agenda-setting, negotiation, and implementation of soft-law commitments. These findings matter for understanding how binding and non-binding international AI governance is shaped and who holds power in shaping it.
- AI policy
Research
Structural Causal World Models for Safety Assurance of AI-based Autonomy
Zou, Jie, STEFANAKOS, IOANNIS, Shahbeigi Roudposhti, Sepeedeh et al.
White Rose Research Online (University of Leeds, The University of Sheffield, University of York) · 2026-06-09
This paper introduces Structural Causal World Models (SCWMs), a formal framework grounded in structural causal models to support safety assurance of AI-based autonomous systems. SCWMs provide interpretable, machine-verifiable representations that unify symbolic constraints, probabilistic uncertainty, and causal dependencies, enabling traceable hazard analysis, safety requirement derivation, and run-time monitoring. The methodology is domain-agnostic and is illustrated through autonomous driving examples, aiming to close the semantic gap in defining safety requirements for complex AI systems. The work contributes to reducing uncertainty in safety assurance by providing a basis for causal hazard and risk analysis and verification of probabilistic guarantees.
- Certifications
- Quality assurance
- AI policy
Research
CCBENCH: Assessing LLM Cultural Competence via Implicitly Signaled Norms using Health Queries
Vasudha Varadarajan, Akhila Yerukola, Mona T. Diab et al.
arXiv · 2026-06-08
CCBENCH is a new benchmarking framework that evaluates how well large language models (LLMs) adapt to users' implicitly signaled cultural values, rather than relying on static demographic labels. Using a health-query case study (CCBENCH-Health), the authors constructed 60 theoretically grounded personas spanning six cultures and assessed five leading LLMs across 3,120 unique interactions. Results show that even the best-performing models give culturally appropriate responses only 20–30% of the time, with chain-of-thought prompting yielding only modest 3–5% average gains. A persistent asymmetry is found where models perform better when personas deviate from cultural norms than when they follow them—most starkly in the Afghan context (average 8.8%)—suggesting models default to built-in biases rather than adapting to cultural cues.
- Quality assurance
- AI policy
Research
SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents
Qianfeng Wen, Yifan Simon Liu, Xin Liu et al.
arXiv · 2026-06-08
SafeGEO examines how Generative Engine Optimization (GEO) — the practice of rewriting web content to boost visibility in AI-generated outputs — can be weaponized in recommendation agents to make flawed products appear better supported than they are. The authors build an evaluation suite with 22 GEO attack variants across 600 recommendation cases and find that such attacks increase the rate at which flawed products enter recommendation sets by up to 83.2%. They also test agent-side defenses like defensive prompting and structured evidence checks, which reduce harmful promotion by up to 39.2%, but cannot fully close the gap to baseline performance without any GEO attacks. The findings highlight a meaningful and unresolved vulnerability in AI-powered recommendation systems where seller-controlled content can systematically mislead AI agents.
- Enterprise
- Quality assurance
Research
Less Context, Better Agents: Efficient Context Engineering for Long-Horizon Tool-Using LLM Agents
Abhilasha Lodha, Mahsa Pahlavikhah Varnosfaderani, Abir Chakraborty et al.
arXiv · 2026-06-08
This paper investigates how to manage context efficiently for LLM-based agents handling enterprise workflows, specifically automated expense itemization in Microsoft Dynamics 365 Finance and Operations. The authors evaluate four GPT-5 configurations on a 50-task hotel expense benchmark, finding that pruning context to the last 5 tool call/response pairs plus automated summarization achieves 91.6% complete itemization and 99.64% average amount itemized, while reducing token usage and runtime compared to retaining full conversation history. Full-context retention achieved only 71.0% completion at nearly 1.5 million tokens and over 14 hours, whereas the pruning-plus-summarization approach used roughly 553,000 tokens and under 6 hours. The results demonstrate that selective retention of recent tool interactions combined with compact summarization improves both reliability and efficiency for enterprise tool-use agent workflows.
- Enterprise
- Quality assurance
Research
Local Is Not a Sufficient Privacy Boundary: Governing OS-Integrated On-Device AI
Jonghyun Chung, Sanket Badhe
arXiv · 2026-06-08
This paper argues that running AI on a local device does not, by itself, constitute a meaningful privacy boundary, because on-device assistants can still aggregate sensitive data from email, calendars, files, and screenshots, persist derived state, invoke tools, and route requests to cloud infrastructure. The authors develop an OS-centered privacy framework that treats privacy as an institutional accountability problem, specifying a threat model, a six-part risk taxonomy, privacy-by-architecture controls, and a four-level audit rubric. They apply the rubric to Apple Intelligence/Foundation Models, Android AICore/Gemini Nano, and Microsoft Recall using publicly available documentation. The work has direct implications for how regulators, platform vendors, and auditors should govern on-device AI, emphasizing constrained information flow, bounded authority, visible user control, and auditable governance across the OS lifecycle.
- AI policy
- Certifications
Research
Gaming AI-Assisted Peer Reviews Poses New Risks to the Scientific Community
Lin Li, Qi Zhang, Xander Davies et al.
arXiv · 2026-06-08
This paper demonstrates that AI-assisted peer review systems are vulnerable to a simple, low-cost adversarial attack: superficially rephrasing a manuscript's abstract without changing its scientific content. The strongest attack achieves a success rate of about 38% in improving AI review outcomes—rising above 50% when the original AI review suggests rejection—increasing acceptance ratings by up to +1.31 points on a 10-point scale for Gemini 3 Flash reviewers and boosting scores on criteria like soundness, significance, and perceived contribution. The attack costs roughly $1 and 5 minutes per submission and is difficult to distinguish from ordinary editing, meaning authors may be incentivized to optimize for AI judgment rather than scientific merit. The authors argue that AI review tools should not be treated as neutral evaluators in high-stakes peer review without systematic robustness testing, transparent safeguards, and careful human oversight.
- Quality assurance
- AI policy
Research
Invisible to humans, visible to machines: a preregistered audit of Unicode fidelity across four biomedical bibliographic APIs
Przemysław Czuma
arXiv · 2026-06-08
This preregistered audit tests whether four major biomedical bibliographic APIs (PubMed E-utilities, Crossref, OpenAlex, Semantic Scholar) faithfully reproduce Unicode characters from published abstracts, using PubMed Central JATS XML as ground truth across a random sample of 4,000 articles. The study finds two systematic, near-total character losses: PubMed preserves typographic punctuation in only 0.6% of eligible abstracts, and OpenAlex preserves special whitespace in 0% of cases, while mathematical symbols and Greek letters are preserved at over 95% fidelity across all APIs. Additionally, Crossref returns no abstract at all for 24.6% of papers, with Elsevier and ACS showing 0% coverage. These findings matter because silently degraded text directly undermines the quality of biomedical LLM training corpora, scientometric analyses, and any corpus-based research that assumes API-returned text matches the published source.
- Quality assurance
Research
A Controlled Audit of Pretraining Contamination in Public Medical Vision-Language Benchmarks
Bruce Changlong Xu, Lan Wu, Alexander Ryu
arXiv · 2026-06-08
This paper audits whether publicly available medical vision-language benchmarks (SLAKE-En, PathVQA, VQA-RAD, and OmniMedVQA) may have been present in the pretraining data of open vision-language models, which would inflate reported accuracy. Using four detection methods—image-side near-neighbour overlap, canonical-order exchangeability, tail enrichment, and cross-model overlap—the authors find measurable image-source overlap on SLAKE-En (up to 19.8% of images flagged) and text-side signals on SLAKE-En and OmniMedVQA, though manual review suggests distributional rather than exact pixel-level duplication. Critically, some detector families (Min-K%++ tail enrichment and cross-model top-K overlap) prove unreliable as standalone contamination signals on small cohorts, as a model without plausible medical-VQA exposure (BLIP-2) reproduces apparent positive signals. The findings raise important concerns about the validity of benchmark evaluations for medical AI and highlight the need for more robust contamination detection methods.
- Quality assurance
- Certifications