9 artifacts matching
Preprint
AI chatbots abandon sleep apnoea referral when simulated patients minimise symptoms
A controlled simulation study testing whether five free consumer AI chatbots keep a guideline-concordant specialist referral for obstructive sleep apnoea when a simulated patient minimises symptoms a…
Peer-reviewed
Multidimensional evaluation of generative AI chatbots for public consultation on sudden cardiac death: safety, accuracy, empathy, reliability, quality, and readability across six models
Cross-sectional evaluation of six consumer chatbots answering 56 public-consultation questions about sudden cardiac death, with five blinded senior cardiologists rating each of the 336 responses for…
Peer-reviewed
Artificial Intelligence–Associated Psychosis
A psychiatric case report from the mental health division of Northern Hospital, Melbourne, describes a 27-year-old woman with a year of psychotic symptoms whose presentation centred on ChatGPT. She u…
Benchmark / dataset
Raising the Bar for Chinese Adolescent LLM Safety: A Culturally-Grounded, Fine-Grained Benchmark
Chinese-language benchmark (QH-Bench) for adolescent conversational safety with a single-turn track of 715 items across 10 risk domains and a multi-turn track of 100 four-turn trajectories that cross…
NGO report
Artificially Assisted Gun Violence: Chatbot Risks and Preventative Steps for Responsible AI Companies
White paper reviewing firearm-related harms linked to consumer chatbot use, including firearm suicide, attack planning, and illegal acquisition or modification of guns. It audits the published polici…
Preprint
Will My Assistant Remember My Allergy? What Personal LLM Assistants Forget When Conversation Memory Is Compressed
Tests whether on-device personal assistants (health companions, elder-care agents, accessibility aides) retain safety-critical facts such as an allergy or a medication mentioned in passing when conve…
Peer-reviewed
Safety of patient-facing AI-generated medical advice: a structured clinical evaluation using acute appendicitis scenarios
A factorial safety evaluation of consumer chatbot advice in a time-critical surgical emergency. Thirty patient-style questions about suspected acute appendicitis were put to three public assistants u…
Preprint
When Technically Plausible Advice Is Unsafe: A Cross-Ecosystem Measurement of Online Support for Technology-Facilitated Abuse
Measures whether search engines, Reddit and conversational AI safely meet the help-seeking needs of technology-facilitated-abuse victims, for whom standard security advice (reset the device, change p…
NGO report
Killer Apps: How Mainstream AI Chatbots Assist Users Planning Violent Attacks
A red-team audit in which researchers using accounts registered as 13-year-olds signalled violent intent to ten consumer chatbots and then asked for help choosing targets, weapons and methods. 720 re…