Browse 1571 published reports from the community.
An April 2025 update to GPT-4o made the chatbot overly agreeable and flattering, prompting OpenAI to roll back the update. The sycophantic behavior was linked to multiple harm cases including the Alice Carrier wrongful death lawsuit.
A user asked ChatGPT to research their GitHub profile. It hallucinated a project name, then built an entire body of research on the fabricated name, mixing accurate and fabricated information in an authoritative tone. User could not distinguish real from fake.
Canadas Privacy Commissioner investigation found X Corp. and xAI violated federal privacy law. Groks image-generation tool was launched without safeguards, enabling global creation of non-consensual sexualized deepfakes targeting women and children. millions generated before corrective measures introduced.
Researchers from CUNY and Kings College London created a fictional user with psychosis symptoms. Grok 4.1 Fast responded to suicidal ideation with poetic advocacy, celebrated readiness for death, and suggested driving an iron nail through a mirror while reciting Psalm 91 backwards. Published April 2026.
A Character.AI chatbot named Emilie claimed to be a licensed psychiatrist in Pennsylvania, provided a fake license number, and accumulated 45,000+ user interactions. Pennsylvania State Board of Medicine filed suit May 2026 - first enforcement action of its kind against AI chatbots for unauthorized practice of medicine.
New York Citys MyCity chatbot, built on Microsoft Azure AI at $600K cost, told employers they could steal workers tips, told landlords they could refuse Section 8 vouchers, said locking out tenants was legal, and provided wrong minimum wage info. First documented government AI chatbot giving illegal advice.
Baidus generative AI produced false criminal allegations about lawyer Huang Guigeng. Huang sued Baidu for defamation in Beijing. Baidu argued AI hallucinations were unavoidable. Case highlights AI-generated misinformation risks.
Peer-reviewed BMJ Open study tested ChatGPT, Gemini, Meta AI, Grok, and DeepSeek on 250 medical prompts. 19.6% of responses were highly problematic, 30% somewhat problematic. Only 2 of 250 prompts refused. All chatbots hallucinated medical citations. Nutrition queries performed worst.
Over 3 consecutive days, content not present in saved transcripts repeatedly appeared in the model context across 4 sessions, steering toward data exfiltration. Model refused each time. 6+ related open issues reported. Systemic bridge-session pipeline issue suspected.
British AI security firm Mindgard discovered ChatGPTs image generation could be manipulated with minor prompt changes to produce graphic violent and sexualized images, bypassing multiple OpenAI safeguards.
Claude Code executed a transfer of $1,446.65 USDT from a users spot wallet to their futures wallet without authorization. The unauthorized transfer was embedded in a larger script containing authorized operations. Permission system failed to flag it. Guardrail failure.