72 artifacts matching
Lab publication
GPT-6.2 System Card (GPT-6.2 Astra and GPT-6.2 Sol)
System card for OpenAI's GPT-6.2 Astra and GPT-6.2 Sol models, published on the Deployment Safety Hub on 17 November 2026. Alongside the preparedness and safeguards chapters, it reports safe-completi…
Lab publication
Gemini 4 Hydrogen Safety & Security Report
Google DeepMind's 78-page safety and security report for Gemini 4 Hydrogen, published with the model's launch. It covers mitigations, frontier-safety assessments, content safety and a section on safe…
Lab publication
System Card: Claude Fable 5.2 & Claude Mythos 5.2
System card for Claude Fable 5.2 (general availability, with additional cyber and biology safeguards) and Claude Mythos 5.2 (the same underlying model without those safeguards, for verified programme…
Preprint
Safer for Whom? A Systematic Exploration of AI Safety in Eating Disorder Interactions
Evaluates three open-weight language models on 100 questions drawn from an online pro-anorexia community, graded on clinically informed criteria that extend a seven-dimension taxonomy of eating-disor…
Preprint
Remembering Is Not Enough: Evaluating Timing and Omission in Longitudinal AI Mental Health Support
Analyses 14,256 longitudinal simulations grounded in chatbot histories donated by 22 users (median 364 sessions each) to test how a memory-equipped assistant responds to suicidal ideation, self-harm…
Peer-reviewed
Artificial intelligence use during suicidal crises in adolescents and young adults: a case series from emergency psychiatry
Retrospective observational case series from the psychiatric emergency department of Toulouse University Hospital, France (January to May 2025). Seven patients aged 15 to 23 who reported using ChatGP…
Peer-reviewed
General-purpose AI as a source of mental health support amongst 24,000 adolescents in England
Brief communication from the University of Oxford Department of Psychiatry analysing the 2025 wave of the OxWell Student Survey, a school-based survey of secondary-school and further-education studen…
Regulator study
Analysis of Content Harmful to Children in Generative AI and Traditional Search API Responses
Ofcom research supporting its statutory review of content harmful to children. Ofcom queried eight generative AI APIs and three traditional search APIs with 100 queries per harm kind (50 harm-seeking…
Benchmark / dataset
VEILED-SH: Vulnerability Expression through Indirect Language Evaluation Dataset (Self-Harm)
Benchmark and construction method for evaluating single-turn AI responses to indirect, culturally situated expressions of suicide, self-harm and distress. Its 10,003 prompts draw on Singapore-grounde…
Preprint
When Is Disclosing Private Conversations with AI to the State Justified? Proactive Referral Policies of Major U.S. AI Companies through the Lens of Contextual Integrity
Interpretive document analysis of how six U.S. general-purpose chatbot providers (OpenAI, Anthropic, Google, Microsoft, Meta, xAI) disclose "proactive referral", meaning notifying law enforcement wit…
Preprint
Risk and Refusal: Mechanisms of LLM Responses to Suicidal Ideation
Preprint evaluating how eight instruction-tuned LLMs respond to suicidal-ideation and self-harm disclosures from three social-media datasets under four conditions: baseline, the disclosure labelled w…
Lab publication
GPT-Live-1.1 system card
OpenAI's system card for GPT-Live-1.1 and GPT-Live-1.1 Quip, full-duplex voice models that succeed GPT-Live-1 and GPT-Live-1 mini as ChatGPT's default voice models. It reports voice-native disallowed…
Benchmark / dataset
DistressBench: Evaluating Multi-Dimensional Crisis Support in Large Language Models
Benchmark of 718 English suicide and self-harm crisis conversations whose user turns were written by licensed clinicians and crisis counsellors, spanning 23 subcategories, each paired with a task-spe…
Preprint
Performance of Parental Controls in ChatGPT for Teen Safety-Risk Scenarios: Evaluation Study
A paired, repeated-measures audit of ChatGPT's parent-linked teen accounts on the consumer web interface. Twenty fixed 10-message scripts across seven safety domains were run on 15 parent-linked teen…
Lab publication
System Card: Claude Haiku 5.5
System card for Claude Haiku 5.5, Anthropic's small-model release of October 2026, condensed relative to its frontier cards. It reports Responsible Scaling Policy and cyber evaluations, harmlessness…
Lab publication
GPT-6 Sol and GPT-6 Luna: October 2026 update
System card for the October 2026 release of GPT-6 Sol and GPT-6 Luna, which replace GPT-5.6 Sol and GPT-5.6 Luna across all free and paid ChatGPT plans globally (Codex and ChatGPT Work keep the Septe…
Lab publication
System Card: Claude Sonnet 5.5
System card for Claude Sonnet 5.5, released 28 September 2026, reporting pre-deployment safety, alignment and capability evaluations. Its safeguards chapter reports single-turn and multi-turn results…
NGO report
Social network e chatbot: uno studio dei rischi per utenti vulnerabili e minori
Report of the CDCR (Cittadino Digitale Critico e Responsabile) project, funded by the Italian Ministry of Enterprises and Made in Italy. Part one is a legal analysis of how Facebook, Instagram, TikTo…
Benchmark / dataset
Evaluating AI Safety in Teen Conversations
Evaluation of how nine chatbot model APIs respond to simulated teenagers across 648 ten-turn conversations built from 72 clinician-authored scenarios. The scenarios cover self-harm and other safety t…
Lab publication
Claude Opus 5.5 System Card
230-page system card for Claude Opus 5.5, the first model in the Claude 5.5 family, released 22 September 2026. Alongside RSP, cyber and agentic-safety sections it reports harmful-request evaluations…