AI Frontier Institute

News & Research

The latest AI research and news with real-world stakes. Each item is sourced, dated, summarized in plain English and tagged by impact area, and checked against its source before it appears.

• ResearcharXiv2026-04-25WP

  How Researchers Navigate Accountability, Transparency, and Trust When Using AI Tools in Early-Stage Research: A Think-Aloud Study · Sanjana Gautam, Houjiang Liu, Yujin Choi et al.

  This think-aloud study with 15 researchers examined how AI tools powered by large language models (LLMs) affect accountability, transparency, and trust during early-stage research tasks such as literature exploration, synthesis, and ideation. The study found that AI outputs' confident tone obscures epistemic uncertainty, making it harder for researchers to identify which outputs need scrutiny; opaque retrieval processes hinder provenance tracing; and trust in AI is fragile and context-dependent. Researchers developed compensatory strategies to restore scholarly judgment when AI tools fell short. The findings frame AI-mediated research as a Responsible AI (RAI) problem and call for deliberate AI integration that preserves accountability and supports informed trust.
• ResearcharXiv2026-04-25QC

  ArgRE: Formal Argumentation for Conflict Resolution in Multi-Agent Requirements Negotiation · Haowei Cheng, Milhan Kim, Chong Liu et al.

  ArgRE is a multi-agent requirements negotiation system that embeds formal argumentation theory (Dung-style abstract argumentation) into the conflict resolution stage, replacing heuristic aggregation with explicit accept/reject decisions traceable at the argument level. Across five case studies in safety-critical, financial, and information-system domains, independent evaluators rated ArgRE's decision justifications significantly higher than heuristic synthesis (4.32 vs. 3.07, p < 0.001), compliance coverage reached 84.7% versus 47.6%–47.8% for baselines, and semantic intent preservation remained comparable (94.9% BERTScore F1). The system also integrates KAOS goal modeling, multi-layer verification, and standards-oriented artifact generation, making it particularly relevant to regulated domains that require auditability. These results demonstrate that principled argumentation semantics can substantially improve the traceability and compliance of AI-assisted requirements engineering.
• ResearcharXiv2026-04-25QP

  Latent Space Probing for Adult Content Detection in Video Generative Models · Alizishaan Khatri, Chiquita Prabhu

  This paper proposes a latent space probing framework for detecting adult and sexually explicit content in AI-generated videos, specifically targeting the CogVideoX video diffusion model. Rather than analyzing text prompts or decoded pixel outputs, the approach intercepts denoised latent representations during inference and attaches lightweight classifiers for real-time detection. The authors build a large-scale binary dataset of 11,039 ten-second video clips and report 97.29% F1 score on a held-out test set with only 4–6ms computational overhead. The results suggest that latent-space signals carry strong discriminative features for harmful content detection, improving both accuracy and efficiency over existing approaches.
• ResearchZenodo (CERN European Organization for Nuclear Research)2026-04-25WP

  The Role of AI and Technology in the Rise of Gig Workers in India: Technological Enablers and Labor Market Impacts · Devesh Lowe

  This paper examines how AI-driven algorithms, mobile applications, cloud computing, and data analytics have enabled India's gig economy to expand, with the gig workforce projected to grow from 7.7 million workers in 2020-2021 to 23.5 million by 2029-2030. Platforms like Swiggy, Zomato, Uber, and Ola have democratized income-generating opportunities and fostered financial inclusion, but have also intensified labor precarity through algorithmic control, income volatility, and inadequate social security. The paper identifies critical policy gaps and recommends frameworks that balance technological innovation with worker welfare in India's rapidly evolving labor market.
• ResearcharXiv (Cornell University)2026-04-25EQCP

  AI Identity: Standards, Gaps, and Research Directions for AI Agents · Takumi Otsuka, Kentaroh Toyoda, Alex Leung

  This paper examines how AI agents operating autonomously across organizational boundaries—running transactions, workflows, and sub-agent chains without continuous human supervision—lack adequate identity and accountability infrastructure. The authors define 'AI Identity' as the continuous relationship between what an agent is declared to be and what it is observed to do, and conduct a gap analysis finding that no current technical or regulatory framework adequately governs these nondeterministic, boundary-crossing entities. They identify five structural gaps—including semantic intent verification, recursive delegation accountability, and governance opacity—that cannot be resolved by engineering effort alone, concluding that foundational research on AI identity is urgently needed. This work has significant implications for enterprise deployment, policy development, certification frameworks, and quality assurance of AI systems.
• ResearchAI & Society2026-04-25QCP

  Explainability gaps in AI-driven criminal justice governance: the RisCanvi case · Oier Mentxaka, Jesus López-Montilla, Dorleta Urrutia-Onate et al.

  This paper examines RisCanvi, an AI-based risk assessment system used in Catalonia to inform imprisonment, conditional release, and rehabilitation decisions. It finds that a 2019 transition from a transparent weighted-sum model to a logistic regression system obscured the variables, weights, and thresholds from professionals and affected individuals, undermining their ability to contest algorithmic decisions. The authors propose a three-layer explainability and governance framework—covering algorithmic explainability, human-centred interface design, and narrative communication—aligned with the EU AI Act and ISO/IEC 42001, aimed at improving auditability, procedural fairness, and accountability in AI-assisted criminal justice.
• ResearchZenodo (CERN European Organization for Nuclear Research)2026-04-25WP

  The Role of AI and Technology in the Rise of Gig Workers in India: Technological Enablers and Labor Market Impacts · Devesh Lowe

  This paper examines how AI-driven algorithms, mobile platforms, cloud computing, and data analytics have fueled India's gig economy, projecting growth from 7.7 million gig workers in 2020-2021 to 23.5 million by 2029-2030. While platforms like Swiggy, Zomato, Uber, and Ola have expanded income-generating opportunities and financial inclusion, the study finds that algorithmic control, income volatility, and inadequate social security have simultaneously deepened labor precarity. The authors identify critical policy gaps and recommend frameworks to balance technological innovation with worker welfare and protections.
• ResearcharXiv2026-04-24CP

  What Should Frontier AI Developers Disclose About Internal Deployments? · Jacob Charnock, Raja Mehta Moreno, Justin Miller et al.

  This paper addresses a governance gap in frontier AI development by proposing a structured disclosure framework for internal AI deployments—cases where highly capable models are used to automate AI R&D within developer organizations. The authors identify four categories of information companies should disclose: capabilities, usage, safety mitigations, and governance, analyzing the benefits, limitations, and risks associated with each type of disclosure. The framework is designed to support both public transparency documents (such as model system cards) and private periodic reports required under emerging frontier AI regulation. The work is directly relevant to policy and certification efforts aimed at ensuring externally verifiable safety evidence for internally deployed frontier models.
• ResearcharXiv2026-04-24QC

  Peer Identity Bias in Multi-Agent LLM Evaluation: An Empirical Study Using the TRUST Democratic Discourse Analysis Pipeline · Juergen Dietrich

  This paper presents the first systematic empirical measurement of peer identity bias in the TRUST multi-agent LLM evaluation pipeline, which exposes model components to each other's identities through multiple structural channels. The central finding is that anonymizing only a single channel produces near-zero apparent bias because individual channels cancel each other out, creating a false impression that identity bias is absent; only full-pipeline anonymization reveals the true pattern, in which homogeneous model ensembles amplify sycophancy while heterogeneous configurations show the reverse. One tested model exhibited baseline sycophancy two to three times higher than others and near-zero deliberative conflict on ideological topics, making it structurally unsuitable for pipelines requiring genuine inter-role disagreement. The authors conclude that systems validated under partial anonymization or homogeneous ensembles may pass validation while retaining hidden structural bias, with direct implications for quality-critical multi-agent LLM deployments.
• ResearcharXiv2026-04-24EP

  Representational Harms in LLM-Generated Narratives Against Global Majority Nationalities · Ilana Nguyen, Harini Suresh, Thema Monroe-White et al.

  This paper investigates how widely-adopted large language models (LLMs) portray individuals from Global Majority nationalities in open-ended narrative generation tasks. The study finds persistent representational harms by national origin, including harmful stereotypes, erasure, and one-dimensional portrayals, with minoritized national identities underrepresented in power-neutral stories and over fifty times more likely to appear in subordinated character roles than dominant ones. These biases are amplified when US nationality cues appear in prompts and persist even when those cues are replaced with non-US identities, ruling out sycophancy as an explanation. The authors warn against uncritical adoption of US-based LLMs for high-stakes applications such as simulated asylum seeker interviews and call for research methodologies that center Global Majority perspectives.
• ResearcharXiv2026-04-24QCP

  Make Mechanistic Interpretability Auditable: A Call to Develop Guidelines via Continuous Collaborative Reviewing · Michael Lan, Narmeen Fatimah Oozeer, Chaithanya Bandi et al.

  This position paper argues that mechanistic interpretability (MI) research lacks the standardized auditing infrastructure needed for adoption in high-stakes applications such as medical AI and autonomous systems. The authors illustrate the problem concretely: conflicting findings across studies on the same neural network behaviors have gone unresolved due to methodological inconsistencies, leaving stakeholders unable to certify the validity of MI results. To address this, the paper proposes a three-part framework — a continuous collaborative reviewing platform, expert-verified guidelines derived from that platform, and source-based auditing that traces how claims depend on prior arguments — designed to complement traditional peer review. The authors contend that auditing MI itself is a prerequisite for its responsible use in AI safety, industry, and governance.
• ResearcharXiv2026-04-24EP

  How Supply Chain Dependencies Complicate Bias Measurement and Accountability Attribution in AI Hiring Applications · Gauri Sharma, Maryam Molamohammadi

  This paper examines how the complex supply chains underlying AI hiring systems—spanning data vendors, model developers, platform providers, and deploying organizations—make it difficult to measure algorithmic bias and assign accountability. The authors show that bias can emerge from interactions among individually compliant components (e.g., a resume parser that is unbiased alone but discriminatory when combined with specific ranking algorithms), and that information asymmetries leave deploying organizations legally responsible for systems they cannot fully inspect. Drawing on literature review and regulatory analysis of frameworks such as the EU AI Act, NYC Local Law 144, and Colorado's AI Act, the paper argues that current technical and regulatory approaches fail to address these fragmented responsibility structures. The authors propose multi-layered interventions including system-level audits, vendor disclosure guidelines, continuous monitoring, and chain-wide documentation to enable meaningful governance of distributed AI development.
• ResearcharXiv2026-04-24WQ

  Rethinking XAI Evaluation: A Human-Centered Audit of Shapley Benchmarks in High-Stakes Settings · Inês Oliveira e Silva, Sérgio Jesus, Iker Perez et al.

  This paper evaluates eight Shapley value variants used in explainable AI (XAI) through a large-scale human-centered study involving professional analysts and 3,735 fraud-detection case reviews. The researchers find a fundamental misalignment: standard quantitative metrics like sparsity and faithfulness do not predict how useful or clear explanations are to human decision-makers. Critically, while no Shapley formulation improved analysts' objective performance, explanations consistently increased their decision confidence—raising a serious risk of automation bias in high-stakes settings. The findings challenge current XAI evaluation practices and provide guidance for selecting explanation methods in operational risk workflows.
• ResearcharXiv2026-04-24WP

  Measuring and Mitigating Persona Distortions from AI Writing Assistance · Paul Röttger, Kobi Hackenburg, Hannah Rose Kirk et al.

  This paper investigates how AI writing assistance distorts the perceived identity, beliefs, and personality of writers—what the authors call 'persona distortions.' Across three large-scale experiments with nearly 3,000 writers and over 11,000 readers, AI-assisted text consistently made writers appear more opinionated, competent, and positive, while shifting perceived demographics toward more privileged groups. Attempts to mitigate these distortions through reward model training were partially successful but reduced user acceptance, suggesting a tension between faithful representation and what users find desirable. The findings carry significant implications for public discourse, political persuasion, and democratic deliberation as AI writing tools scale to hundreds of millions of users.
• ResearcharXiv2026-04-24WP

  Smaller, Younger, and More Impactful: How AI-Assisted Writing Transforms Research Teams · Haoyang Wang, Mingze Zhang, Yi Bu et al.

  Analyzing 147,074 full-text publications from the PLoS family and Nature portfolio since 2020, this study finds that research teams using AI-assisted writing tend to be smaller and composed of younger researchers compared to those that do not. Crucially, these more compact, junior-leaning teams show a higher probability of producing highly impactful publications, suggesting AI writing tools can offset traditional advantages of large, senior teams. The authors use multiple statistical methods—including propensity score matching—to support these findings, and conclude that the trend warrants policy changes in research evaluation, funding, and training.
• ResearcharXiv2026-04-24WQ

  Learning-augmented robotic automation for real-world manufacturing · Yunho Kim, Quan Nguyen, Taewhan Kim et al.

  This paper introduces Learning-Augmented Robotic Automation (LARA), a hybrid system combining learned task controllers and a neural 3D safety monitor with conventional industrial robot workflows. Deployed on a live electric-motor production line, the system automated deformable cable insertion and soldering — tasks previously done manually — using less than 20 minutes of real-world training data per task. Over a continuous 5-hour 10-minute run producing 108 motors, it achieved a 99.4% pass rate on product-level quality-control tests, maintained near-human takt time, and reduced variability in solder-joint quality and cycle time without physical safety fencing. The results demonstrate that learning-based robot control can meet industrial standards for reliability, quality, and human safety outside of laboratory settings.
• ResearcharXiv2026-04-24QP

  Calibration Drift Under Reasoning: How Chain-of-Thought Budgets Induce Overconfidence in Large Language Models · Prakul Sunil Hiremath, Harshit R. Hiremath

  This paper investigates how chain-of-thought (CoT) reasoning affects the calibration of large language models (LLMs), finding that extending reasoning beyond a task-specific threshold can cause systematic overconfidence — a phenomenon the authors term Calibration Drift Under Reasoning (CDUR). Using Llama-3.1-8B and Llama-3.3-70B evaluated on 47 reasoning-trap questions across four reasoning budgets and three seeds (1,368 API calls; 574 valid responses), they show that Expected Calibration Error follows a non-monotonic pattern: it first decreases as reasoning corrects errors, then increases as longer reasoning produces internally consistent but incorrect explanations. To address this, the authors propose CABStop, a calibration-aware stopping rule that halts reasoning when confidence diverges from an auxiliary accuracy estimate. These findings indicate that deeper reasoning does not reliably improve model reliability, which has direct implications for the safe deployment of LLMs in high-stakes settings.
• ResearcharXiv2026-04-24QC

  An LLM-Native Psychometric Instrument Reveals a Self-Report--Behavior Gap Across 25 Models · Juan Manuel Contreras

  This paper investigates whether the well-known gap between LLMs' self-reported personality traits and their actual behavior is simply an artifact of applying human-derived psychological categories to models. The researchers built the first psychometric instrument grounded in LLM behavior itself, administering 300 items to 25 LLMs across 17 model families and identifying five reliable behavioral factors (Responsiveness, Deference, Boldness, Guardedness, and Verbosity). Even with these LLM-native constructs, self-report still failed to predict how models actually behaved as rated by 151 human observers, showing the self-report–behavior gap is not merely a category-mismatch problem. Critically, the study finds that LLM self-report items and LLM judges share a source of variance invisible to human raters—a confound that standard within-ensemble reliability checks cannot detect, posing concrete risks for LLM-as-judge evaluation pipelines widely used in model assessment.
• ResearcharXiv2026-04-24QP

  Behavioral Canaries: Auditing Private Retrieved Context Usage in RL Fine-Tuning · Chaoran Chen, Dayu Yuan, Peter Kairouz

  This paper introduces 'Behavioral Canaries,' an auditing framework designed to detect unauthorized use of legally protected retrieved documents in Reinforcement Learning Fine-Tuning (RLFT) of large language models. Standard auditing methods like verbatim memorization checks and membership inference fail for RL-trained models because RL shapes behavioral style rather than retaining specific facts. The proposed framework plants specially crafted preference data—pairing document triggers with distinctive stylistic feedback—so that if those documents are used in training, a detectable behavioral signal emerges. Empirical results show a 67% detection rate at a 10% false-positive rate (AUROC = 0.756) at a 1% canary injection rate, establishing behavioral canaries as a viable tool for auditing training-time influence that manifests as distributional behavioral change.
• ResearcharXiv2026-04-24QP

  Estimating Tail Risks in Language Model Output Distributions · Rico Angell, Raghav Singhal, Zachary Horvitz et al.

  This paper addresses a critical gap in AI safety evaluation: current methods focus on what inputs cause harmful outputs but ignore how probable those harmful outputs actually are. The authors propose an importance-sampling approach that creates 'unsafe' versions of a target language model to efficiently estimate the probability of harmful outputs, achieving 10–20x sample efficiency gains over brute-force Monte Carlo methods and enabling probability estimates as low as 10^-4 with just 500 samples. The method also reveals model sensitivity to input perturbations and can predict deployment risks, making it practically useful for real-world safety assessments. Because language models are queried billions of times daily, even very rare harmful behaviors will occur frequently in aggregate, making accurate tail-risk estimation essential for responsible deployment.
• ResearcharXiv2026-04-24QP

  PrivSTRUCT: Untangling Data Purpose Compliance of Privacy Policies in Google Play Store · Bhanuka Silva, Anirban Mahanti, Aruna Seneviratne et al.

  PrivSTRUCT is a new encoder-decoder framework that analyzes privacy policies in Android apps by preserving the document's logical hierarchy—such as section headings—rather than treating the text as flat, uniform content. Applied to 3,756 Google Play Store apps, it extracts more than twice as many data item and purpose excerpts compared to the leading tool PoliGrapher. The study reveals a critical transparency gap: developers who rely on globally defined purposes rather than locally scoped disclosures are 20.4% more likely to overstate data purposes for first-party collection and 9.7% more likely for third-party sharing. Sensitive data flows, such as sharing financial data for analytics, are frequently obscured under generic categories, pointing to systemic failures in how app developers disclose data practices.
• ResearcharXiv2026-04-24Q

  Reliable Self-Harm Risk Screening via Adaptive Multi-Agent LLM Systems · Meghana Karnam, Ananya Joshi

  This paper presents a statistical framework for multi-agent LLM pipelines used in behavioral health screening tasks such as self-harm risk assessment. The framework models each agent as a stochastic categorical decision and introduces tighter confidence bounds, a bandit-based adaptive sampling strategy, and logarithmic regret guarantees for multi-agent systems. Evaluated on two labeled behavioral health datasets (AEGIS 2.0, N=161; SWMH Reddit posts, N=250), the adaptive sampling approach reduces false positive rates by roughly 40% compared to single-agent models (0.095 vs. 0.159 on AEGIS 2.0) without sacrificing recall. These results suggest that principled adaptive decision-making can meaningfully improve reliability in safety-critical AI screening applications.
• ResearcharXiv2026-04-24P

  When AI Speaks, Whose Values Does It Express? A Cross-Cultural Audit of Individualism-Collectivism Bias in Large Language Models · Pruthvinath Jeripity Venkata

  This study audited three major AI systems (Claude Sonnet 4.5, GPT-5.4, and Gemini 2.5 Flash) for cultural value bias by presenting them with personal dilemmas framed for users across 10 countries, 5 continents, and 7 languages (840 scored responses), then comparing the advice against World Values Survey Wave 7 data on what people in each country actually believe. All three systems consistently gave Western-style, individualist advice even to users from collectivist societies, with a statistically significant mean gap of +0.76 on a 1–5 scale (t=15.65, p<0.001) and gaps as large as +1.85 for Nigeria. The models also differed in mechanism—Claude shifts toward collectivism in native-language prompts, Gemini shifts more individualist, and GPT-5.4 responds only to stated country identity—while Japan was over-stereotyped as more group-oriented than surveys indicate, revealing encoding of outdated cultural assumptions. The findings suggest frontier AI systems are systematically homogenizing human values toward a Western-individualist baseline, with implications for policy governing AI deployment across diverse populations.
• ResearchAmerican Journal of Health-System Pharmacy2026-04-24ECP

  Balancing opportunities and risks of artificial intelligence in drug policy and regulation · Pineal Bareamichael, Tinglong Dai, Mariana P. Socal et al.

  This paper examines how artificial intelligence is reshaping five domains of prescription drug policy—drug discovery, regulatory efficiency, coverage decisions, pricing, and supply chains—while also raising concerns about access, equity, and ethical risks. The authors highlight that AI-designed drug candidates have shown early promise, such as one therapy entering phase 2 trials in roughly one-third the conventional development time, though no fully AI-designed drug has yet reached market approval. Without deliberate policy guidance and transparency requirements, the authors warn that AI could reinforce existing inequities and shift priorities away from public health needs. Pharmacists are identified as key advocates for ensuring AI tools remain transparent, equitable, and patient-centered.
• ResearcharXiv (Cornell University)2026-04-24EQP

  The Security Cost of Intelligence: AI Capability, Cyber Risk, and Deployment Paradox · Sukwoong Choi

  This paper develops an analytical model examining how firms decide on AI deployment and cybersecurity investment when organizational governance controls have not kept pace with AI capability. The central finding is a 'deployment paradox': in high-loss environments, more capable AI can actually lead firms to deploy less, because greater capability requires broader authority exposure (data access, delegated workflows) that increases breach risk under weak governance. The model shows optimal deployment falls below a no-risk benchmark, with the gap worsening as breach-loss magnitude and authority exposure grow, and that governance maturity is a prerequisite—not just a constraint—for translating AI capability into productive use. This has direct implications for enterprise AI adoption strategies and the design of organizational and policy frameworks around AI governance.

← Back to the dashboard