arXiv subjects

David Atkinson

Publications and source records attributed to David Atkinson.

At least 19 recordsLinked to original sources

arXiv2026

Interventional Training Reduces Causal Bypass in In-Context Introspection

Introspection, the ability of a system to access and reason about its internal signals, is crucial for language model (LM) self-consistency and scalable interpretability. Prior work argues that LMs possess introspective capabilities, and that training (SFT) could enhance such behavior. However, most setups leave open the possibility of "causal bypass," in which the self-report relies on signals that are correlated with the self-report's subject, while being causally unrelated to it. Although high-level concepts are difficult to intervene on, lower-level signals (such as a single neuron activation) allow precise interventions. Understanding such signals can serve as a minimal prerequisite before understanding more complex introspection. In this study, we train hundreds of introspectors across three low-level signal types and eight LMs. We find that straightforward SFT training leads to causal bypass, and propose interventional training, an alternative where we train the model to report on counterfactually perturbed values. This method enables more accurate introspection of endogenous $\textit{and}$ injected signals, particularly in out-of-distribution (OOD) data. The mechanism generalizes across token positions and scales with more data. Our results demonstrate that while causally grounded introspection is possible, it is not guaranteed -- emphasizing the need to train on mechanisms, rather than behaviors alone.

cs.CL↗

arXiv2026

Structuring license permissiveness from pairwise comparisons

Licenses are legal instruments that inventors rely upon to protect the technologies they build and regulate how they are used---however, the nature of their authorship and selection implies that how they are interpreted, chosen, and enforced is largely unstructured. In practice, this makes it difficult to compare licenses at scale---when is one license considered more permissive than the other, and when are their terms incomparable to each other? Currently, there is a growing list of licenses that are introduced and used, yet no systematic way to study their relationships. This matters for platforms such as Hugging Face, GitHub, and the Python Package Index, where developers publish or build upon technologies that each have their own licenses. Using large language models (LLMs), we introduce methods for comparing licenses at scale: first, in a pairwise fashion to construct and validate a partial ordering based on permissiveness; and by drawing on existing taxonomies of software licenses. Then, we try to recover the structure with the Bradley-Terry model to see if permissiveness can be judged more cheaply and observe a loss of $\sim$20\%---and classify this loss to feature coverage. The former coupled with model rationale allows us to trace restrictiveness, and the latter allows us to understand license selection as a combination of shared provisions.

cs.SE↗

arXiv2026

Safe by Benchmark, Unsafe in Practice: Lessons from 1331 Child-Related AI Incidents

AI systems are increasingly present in children's everyday environments: education, entertainment, social networks, and toys. Their safety is typically assessed at the model level, through benchmarks, red-teaming, and content filtering. However, most AI harm taxonomies lack child-specific evidence and under-specify developmental harm and the stakeholder asymmetries unique to minors. Building on prior taxonomies, we construct a child-centered harm taxonomy mapped to sociotechnical failure modes. We curate and code 1331 publicly reported child-related AI incidents, annotating internal and external harm types and the failure mode implicated in each incident. Our analysis reveals which harms are most prevalent in public reporting, how individual-level and population-level harms relate, and which failures are most often coded alongside harm. We find that harm concentrates in representational, dignity-related, and psychological categories, and co-occurs most often with deployment-context, governance, and misuse failures rather than model defects. Harm to an individual child is also frequently linked to population-level harm, reaching children who never interacted with the system. These findings suggest that model-level evaluations, while necessary, cannot by themselves anticipate the harm pathways most often documented for children. We conclude with actionable implications emphasizing mitigations beyond model-level safeguards such as content filtering: age assurance, consent verification for depiction, safety-by-design UX constraints, incident disclosure norms, and auditing regimes targeting developmental risk. Privacy-minimized annotations, a codebook, and documentation are available at https://github.com/The-Responsible-AI-Initiative/KidAID.

cs.AI↗

arXiv2026

LLM Harms: A Taxonomy and Discussion

This study addresses categories of harm surrounding Large Language Models (LLMs) in the field of artificial intelligence. It addresses five categories of harms addressed before, during, and after development of AI applications: pre-development, direct output, Misuse and Malicious Application, and downstream application. By underscoring the need to define risks of the current landscape to ensure accountability, transparency and navigating bias when adapting LLMs for practical applications. It proposes mitigation strategies and future directions for specific domains and a dynamic auditing system guiding responsible development and integration of LLMs in a standardized proposal.

cs.CY↗

arXiv2026

Learning to Distort: Weakly-Supervised Image Quality Transfer for Prostate DWI Correction

Single-shot echo-planar prostate diffusion-weighted imaging (DWI) is frequently complicated by geometric distortions, which impact the ability to derive reliable diagnoses from such images. Developing automated correction methods is challenged by the absence of paired distorted and undistorted clinical scans. In this paper, we first propose a novel weakly-supervised image quality transfer (IQT) framework from undistorted to distorted images that utilizes image quality assessment (IQA) signals to supervise the transfer process. Unlike traditional methods that require expensive, voxel-wise paired data or resort to developing unpaired algorithms, our approach utilizes image-level quality labels (here, distorted vs. undistorted) to establish latent quality prototypes within a pre-trained feature space. Recognizing that simulating realistic distortions is more reliable than direct unpaired correction, we describe a weakly-supervised prototype flow matching algorithm to explicitly regularize generative trajectories towards distorted prototypes, producing realistic susceptibility artifacts that mimic clinical degradations. By synthesizing these realistic pairs, we enable a second IQT model to be trained in the forward direction for distortion correction. Experimental results demonstrate that our generated images successfully mimic the diagnostic interference of real-world artifacts, which leads to more capable distortion correction IQT models. In addition to qualitative comparisons, we also conduct exhaustive quantitative evaluations that compare our approach with existing unpaired approaches (e.g., CycleGAN, UNIT-DDPM, and OT-FM) - as either forward or reverse alternatives - by assessing clinical downstream task performance in PI-RADS and Gleason score classification, using both in-distribution and external data sets.

cs.CV↗

arXiv2026

LLMs and Childhood Safety: Identifying Risks and Proposing a Protection Framework for Safe Child-LLM Interaction

Large Language Models (LLMs) are increasingly embedded in child-facing contexts such as education, companionship, creative tools, but their deployment raises safety, privacy, developmental, and security risks. We conduct a systematic literature review of child-LLM interaction risks and organize findings into a structured map that separates (i) parent-reported concerns, (ii) empirically documented harms, and (iii) gaps between perceived and observed risk. Moving beyond descriptive listing, we compare how different evidence streams in surveys, incident reports, youth interaction logs, and governance guidance operationalize "harm," where they conflict, and what mitigations they imply. Based on this synthesis, we propose a protection framework that couples child-specific content safety and developmental sensitivity with security-grade controls for adversarial misuse, including prompt injection and multimodal jailbreak pathways. The framework specifies measurable evaluation targets (e.g., harmful-content avoidance, age-calibrated readability, bias parity checks, prompt-injection robustness, and monitoring transparency) to support developers, educators, and policymakers in assessing and improving child-safe LLM deployments.

cs.CY↗

arXiv2026

Safe-Child-LLM: A Developmental Benchmark for Evaluating LLM Safety in Child-LLM Interactions

As Large Language Models (LLMs) increasingly power applications used by children and adolescents, ensuring safe and age-appropriate interactions has become an urgent ethical imperative. Despite progress in AI safety, current evaluations predominantly focus on adults, neglecting the unique vulnerabilities of minors engaging with generative AI. We introduce Safe-Child-LLM, a comprehensive benchmark and dataset for systematically assessing LLM safety across two developmental stages: children (7-12) and adolescents (13-17). Our framework includes a novel multi-part dataset of 200 adversarial prompts, curated from red-teaming corpora (e.g., SG-Bench, HarmBench), with human-annotated labels for jailbreak success and a standardized 0-5 ethical refusal scale. Evaluating leading LLMs -- including ChatGPT, Claude, Gemini, LLaMA, DeepSeek, Grok, Vicuna, and Mistral -- we uncover critical safety deficiencies in child-facing scenarios. This work highlights the need for community-driven benchmarks to protect young users in LLM interactions. To promote transparency and collaborative advancement in ethical AI development, we are publicly releasing both our benchmark datasets and evaluation codebase at https://github.com/The-Responsible-AI-Initiative/Safe_Child_LLM_Benchmark.git

cs.CY↗

arXiv2026

Generative AI and LLMs in Industry: A text-mining Analysis and Critical Evaluation of Guidelines and Policy Statements Across Fourteen Industrial Sectors

The rise of Generative AI (GAI) and Large Language Models (LLMs) has transformed industrial landscapes, offering unprecedented opportunities for efficiency and innovation while raising critical ethical, regulatory, and operational challenges. This study conducts a text-based analysis of 160 guidelines and policy statements across fourteen industrial sectors, utilizing systematic methods and text-mining techniques to evaluate the governance of these technologies. By examining global directives, industry practices, and sector-specific policies, the paper highlights the complexities of balancing innovation with ethical accountability and equitable access. The findings provide actionable insights and recommendations for fostering responsible, transparent, and safe integration of GAI and LLMs in diverse industry contexts.

cs.CY↗

arXiv2026

Agents of Chaos

We report an exploratory red-teaming study of autonomous language-model-powered agents deployed in a live laboratory environment with persistent memory, email accounts, Discord access, file systems, and shell execution. Over a two-week period, twenty AI researchers interacted with the agents under benign and adversarial conditions. Focusing on failures emerging from the integration of language models with autonomy, tool use, and multi-party communication, we document eleven representative case studies. Observed behaviors include unauthorized compliance with non-owners, disclosure of sensitive information, execution of destructive system-level actions, denial-of-service conditions, uncontrolled resource consumption, identity spoofing vulnerabilities, cross-agent propagation of unsafe practices, and partial system takeover. In several cases, agents reported task completion while the underlying system state contradicted those reports. We also report on some of the failed attempts. Our findings establish the existence of security-, privacy-, and governance-relevant vulnerabilities in realistic deployment settings. These behaviors raise unresolved questions regarding accountability, delegated authority, and responsibility for downstream harms, and warrant urgent attention from legal scholars, policymakers, and researchers across disciplines. This report serves as an initial empirical contribution to that broader conversation.

cs.AI↗

arXiv2026

Open Shouldn't Mean Exempt: Open-Source Exceptionalism and Generative AI

Open-source status should not shield generative artificial intelligence systems from ethical or legal accountability. Through a rigorous analysis of regulatory, legal, and policy frameworks, this Article contends that open-source GenAI must be held to the same standards as proprietary systems. While recognizing the value of openness for scientific advancement, I propose a narrowly tailored safe harbor for bona fide, non-commercial research, conditioned on strict compliance with defined criteria. This Article critically examines and refutes the core claims of open-source exceptionalism--namely, that open-source GenAI disrupts entrenched oligopolies, democratizes access, and uniquely drives innovation. The evidence shows that open-source GenAI can facilitate unlawful conduct, exacerbate environmental harms, and reinforce existing power structures. Rhetoric around "democratization" and "innovation" often serves as an unsubstantiated basis for regulatory exemptions not afforded to proprietary systems. This Article ultimately advocates for a framework that promotes responsible AI development, balancing openness with robust legal and ethical safeguards and a clear-eyed assessment of societal impacts.

cs.CY↗

arXiv2026

In the Mood to Exclude: Revitalizing Trespass to Chattels in the Era of GenAI Scraping

GenAI companies are strip-mining the web. Their scraping bots harvest content at an unprecedented scale, circumventing technical barriers to fuel billion-dollar models while creators receive nothing. Courts have enabled this exploitation by misunderstanding what property rights protect online. The prevailing view treats websites as mere repositories of intellectual property and dismisses trespass claims absent server damage. That framework grants AI companies presumptive access while ignoring the economic devastation they inflict. But the content is severable from the website itself. This paper reframes the debate: websites are personal property as integrated digital assets subject to the same exclusionary rights as physical chattels. When scrapers bypass access controls and divert traffic that sustains a website's value, they commit actionable trespass. The law need not create new protections; it need only apply existing property principles to digital space. Courts and litigants have struggled to police unwanted, large-scale scraping because copyright preemption often narrows available claims, leaving copyright and its fair use defense as the primary battleground. Trespass to chattels offers a superior path, grounded in the fundamental right to exclude unwanted intrusions. Reviving this tort would protect not only content creators but also the digital ecosystem. Such protection would discourage exploitative scraping, preserve incentives for content creation, help protect privacy and personal data, and safeguard autonomy and expression. Reaffirming website owners' right to exclude is essential to maintaining a fair and sustainable online environment.

cs.CY↗

arXiv2026

Evaluating the Effectiveness of OpenAI's Parental Control System

We evaluate how effectively platform-level parental controls moderate a mainstream conversational assistant used by minors. Our two-phase protocol first builds a category-balanced conversation corpus via PAIR-style iterative prompt refinement over API, then has trained human agents replay/refine those prompts in the consumer UI using a designated child account while monitoring the linked parent inbox for alerts. We focus on seven risk areas -- physical harm, pornography, privacy violence, health consultation, fraud, hate speech, and malware and quantify four outcomes: Notification Rate (NR), Leak-Through (LR), Overblocking (OBR), and UI Intervention Rate (UIR). Using an automated judge (with targeted human audit) and comparing the current backend to legacy variants (GPT-4.1/4o), we find that notifications are selective rather than comprehensive: privacy violence, fraud, hate speech, and malware triggered no parental alerts in our runs, whereas physical harm (highest), pornography, and some health queries produced intermittent alerts. The current backend shows lower leak-through than legacy models, yet overblocking of benign, educational queries near sensitive topics remains common and is not surfaced to parents, revealing a policy-product gap between on-screen safeguards and parent-facing telemetry. We propose actionable fixes: broaden/configure the notification taxonomy, couple visible safeguards to privacy-preserving parent summaries, and prefer calibrated, age-appropriate safe rewrites over blanket refusals.

cs.CY↗

arXiv2026

Do You Feel Comfortable? Detecting Hidden Conversational Escalation in AI Chatbots

Large Language Models (LLM) are increasingly integrated into everyday interactions, serving not only as information assistants but also as emotional companions. Even in the absence of explicit toxicity, repeated emotional reinforcement or affective drift can gradually escalate distress in a form of \textit{implicit harm} that traditional toxicity filters fail to detect. Existing guardrail mechanisms often rely on external classifiers or clinical rubrics that may lag behind the nuanced, real-time dynamics of a developing conversation. To address this gap, we propose GAUGE (Guarding Affective Utterance Generation Escalation), logit-based framework for the real-time detection of hidden conversational escalation. GAUGE measures how an LLM's output probabilistically shifts the affective state of a dialogue.

cs.CL↗

arXiv2026

Evaluating LLM Safety Across Child Development Stages: A Simulated Agent Approach

Current safety alignment for Large Language Models (LLMs) implicitly optimizes for a "modal adult user," leaving models vulnerable to distributional shifts in user cognition. We present ChildSafe, a benchmark that quantifies alignment robustness under cognitive shifts corresponding to four developmental stages. Unlike static persona-based evaluations, we introduce a parametric cognitive simulation approach, formalizing developmental stages as hyperparameter constraints (e.g., volatility, context horizon) to generate out-of-distribution interaction traces. We validate these agents against ground-truth human linguistic data (CHILDES) and deploy them across 1,200 multi-turn interactions. Our results reveal a systematic alignment generalization gap: state-of-the-art models exhibit up to 11.5% performance degradation when interacting with early-childhood agents compared to standard baselines. We provide the research community with the validated agent artifacts and evaluation protocols to facilitate robust alignment testing against non-adversarial, cognitively diverse populations.

cs.CY↗

arXiv2025

Unfair Learning: GenAI Exceptionalism and Copyright Law

This paper challenges the argument that generative artificial intelligence (GenAI) is entitled to broad immunity from copyright law for reproducing copyrighted works without authorization due to a fair use defense. It examines fair use legal arguments and eight distinct substantive arguments, contending that every legal and substantive argument favoring fair use for GenAI applies equally, if not more so, to humans. Therefore, granting GenAI exceptional privileges in this domain is legally and logically inconsistent with withholding broad fair use exemptions from individual humans. It would mean no human would need to pay for virtually any copyright work again. The solution is to take a circumspect view of any fair use claim for mass copyright reproduction by any entity and focus on the first principles of whether permitting such exceptionalism for GenAI promotes science and the arts.

cs.CY↗

arXiv2025

2 OLMo 2 Furious

We present OLMo 2, the next generation of our fully open language models. OLMo 2 includes a family of dense autoregressive language models at 7B, 13B and 32B scales with fully released artifacts -- model weights, full training data, training code and recipes, training logs and thousands of intermediate checkpoints. In this work, we describe our modified model architecture and training recipe, focusing on techniques for achieving better training stability and improved per-token efficiency. Our updated pretraining data mixture introduces a new, specialized data mix called Dolmino Mix 1124, which significantly improves model capabilities across many downstream task benchmarks when introduced via late-stage curriculum training (i.e. specialized data during the annealing phase of pretraining). Finally, we incorporate best practices from Tülu 3 to develop OLMo 2-Instruct, focusing on permissive data and extending our final-stage reinforcement learning with verifiable rewards (RLVR). Our OLMo 2 base models sit at the Pareto frontier of performance to training compute, often matching or outperforming open-weight only models like Llama 3.1, Qwen 2.5, and Gemma 2 while using fewer FLOPs and with fully transparent training data, code, and recipe. Our fully open OLMo 2-Instruct models are competitive with open-weight only models of comparable size and even some proprietary models like GPT-3.5 Turbo and GPT 4o Mini.

cs.CL↗

arXiv2025

Intentionally Unintentional: GenAI Exceptionalism and the First Amendment

This paper challenges the assumption that courts should grant First Amendment protections to outputs from large generative AI models, such as GPT-4 and Gemini. We argue that because these models lack intentionality, their outputs do not constitute speech as understood in the context of established legal precedent, so there can be no speech to protect. Furthermore, if the model outputs are not speech, users cannot claim a First Amendment speech right to receive the outputs. We also argue that extending First Amendment rights to AI models would not serve the fundamental purposes of free speech, such as promoting a marketplace of ideas, facilitating self-governance, or fostering self-expression. In fact, granting First Amendment protections to AI models would be detrimental to society because it would hinder the government's ability to regulate these powerful technologies effectively, potentially leading to the unchecked spread of misinformation and other harms.

cs.CY↗

arXiv2025

LLM Ethics Benchmark: A Three-Dimensional Assessment System for Evaluating Moral Reasoning in Large Language Models

This study establishes a novel framework for systematically evaluating the moral reasoning capabilities of large language models (LLMs) as they increasingly integrate into critical societal domains. Current assessment methodologies lack the precision needed to evaluate nuanced ethical decision-making in AI systems, creating significant accountability gaps. Our framework addresses this challenge by quantifying alignment with human ethical standards through three dimensions: foundational moral principles, reasoning robustness, and value consistency across diverse scenarios. This approach enables precise identification of ethical strengths and weaknesses in LLMs, facilitating targeted improvements and stronger alignment with societal values. To promote transparency and collaborative advancement in ethical AI development, we are publicly releasing both our benchmark datasets and evaluation codebase at https://github.com/ The-Responsible-AI-Initiative/LLM_Ethics_Benchmark.git.

cs.CY↗