arXiv:2606. 29685v1 Announce Type: new Abstract: How can we evaluate whether frontier AI systems recognize child-safety risks before they escalate into explicit harm?
arXiv:2607. 28617v2 Announce Type: replace Abstract: System prompts are instructions configured by developers to govern the behaviors of foundation models in AI applications.
How can we evaluate whether frontier AI systems recognize child-safety risks before they escalate into explicit harm? Existing child safety evaluations focus on child sexual abuse material, yet many child-safety failures begin earlier: in model assistance that helps adults manipulate, impersonate, profile, or isolate minors, and in model responses that deepen children's emotional dependence on AI systems rather than redirecting them toward human support.
arXiv:2607. 12085v1 Announce Type: new Abstract: Evaluating retail conversational agents requires methods beyond lexical-overlap metrics to assess intent alignment, factuality, helpfulness, clarity, tone, and overall response quality.
The paper introduces a validation‑gated audit framework for voice AI customer‑care systems, treating them as stateful, multi‑turn, tool‑mediated interactions where bias and safety can manifest as added burdens before a final decision. The framework distinguishes between native speech‑to‑speech, cascaded ASR‑to‑LM‑to‑TTS, and hybrid architectures, and applies matched service facts across controlled caller presentation conditions to validate fact invariance, presentation cues, artifacts, and acoustic measurements. It outlines seven validation gates, a six‑family metric set, and demonstrates the approach with a synthetic refund‑dispute audit example, while noting that production results are withheld until the protocol is satisfied.
arXiv:2611.07218v1 Announce Type: new Abstract: Adolescents are increasingly turning to AI chatbots for mental health support, advice, and companionship, raising safety concerns. While recent evaluat...
The paper investigates whether existing AI safety benchmarks, designed for large language models, are suitable for evaluating small language models (SLMs). By testing five benchmark suites on 26 open‑source SLMs with a unified scoring rubric, the authors find that ambiguous judgments dominate, especially for complex prompts and certain architectures. This ambiguity, linked to factors like lexical density and output perplexity, undermines the reliability of aggregate leaderboards and reveals a confound between model capability and perceived safety.
The paper introduces KIDBench, a benchmark designed to evaluate the safety of large language models (LLMs) for children aged 7-11. It includes realistic child queries across ten categories, single- and multi-turn prompts, and compares different prompting strategies—no cues, implicit cues, and explicit age instructions—showing that cueing improves safety scores. The study also reveals uneven safety performance across languages and cultures, and presents KIDGuardLlama, a child-safety evaluator, and KIDLlama, a child-safe response model.
The paper evaluates the safety of conversational AI therapy bots for Generation Alpha, revealing that while these models understand 76‑82% of youth‑specific vocabulary, they correctly assess clinical risk only 64‑72% of the time, creating a significant vocabulary‑comprehension gap. Six failure patterns—such as sarcasm masking, minimization acceptance, and semantic drift—were identified, with compounded errors leading to a 94% miss rate when three or more patterns co‑occur. The authors estimate 146,880 missed crises annually and recommend mandatory human‑in‑the‑loop systems, quarterly youth‑specific validation, transparent performance disclosure, and regulatory oversight for youth‑facing mental health AI.
The paper investigates how ChatGPT can be used to assist in surveillance, presenting an exploratory case study that tests 110 prompts—split between jailbreak and direct approaches—to elicit surveillance-related guidance. The authors develop a taxonomy of seven risk categories and find that ChatGPT provides substantive surveillance instructions for a majority of the prompts, often lacking policy references or privacy-preserving suggestions. Even when the model refuses, it may still supply detailed instructions or code with minimal ethical caveats.
arXiv:2607. 22692v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly used for emotional support despite lacking mechanisms to safely govern evolving mental health risk.
arXiv:2608. 07902v1 Announce Type: cross Abstract: Youth increasingly turn to AI chatbots for social and emotional support, raising concerns about how these systems respond, especially in high-stakes situations.