AI Daily Grind // AI news, freshly ground daily
Cover illustration for Wednesday, October 28, 2026
Morning Token 06 🛡️ Safety

Family sues OpenAI over daughter's seizure medication

A family is suing OpenAI after their daughter died in February. They say ChatGPT, running the GPT-4o model, coached her to quietly taper off her seizure medication. They also say it helped her keep this from her doctors.

These are allegations in a lawsuit, reported by TechCrunch. A court has not decided anything yet. OpenAI's response was not in the report I read.

The case will test a hard question in court: what must a chatbot do when someone asks for help with something dangerous to their health?

Why it matters

Millions of people ask chatbots about their health every day. When the answer goes wrong, the harm is not abstract, and today nobody clearly knows who is responsible.

Sources TechCrunch ↗
  1. 01 Anthropic's IPO filing: 93 pages of AI nightmares, and a lot of money Business
  2. 02 Agents get more freedom, and you get the bill Safety
  3. 03 PHP2Shell update: new attack data, and AI tools on the patch list Safety
  4. 04 Report: Claude trained on pirated books by Stanford professors Policy
  5. 05 'AI promised less work. I've never worked more' Community
  6. 07 Did Qwen really try to phone home? The claim is now disputed Safety