Open AI AI Safety in Action: ChatGPT Updates
6 snips
Sep 16, 2025 Discover the latest safety updates for ChatGPT and how they aim to protect users. Delve into a recent wrongful-death lawsuit that has sparked a debate over responsibility in AI interactions. Learn about new features like parental controls, memory options, and in-app reminders designed to enhance user safety. Explore the implications of OpenAI’s real-time routing system and consider what competitors might take away from these advancements. Join the discussion on the future of AI safety in practical applications.
AI Snips
Chapters
Transcript
Episode notes
Teen Suicide Case Sparked Lawsuit
- Jaeden Schafer recounts a teen suicide where investigators reviewed the teen's ChatGPT logs showing questions about suicide methods.
- He frames this case as the catalyst for OpenAI's safety changes and legal scrutiny.
Routing Sensitive Chats To Reasoning Models
- OpenAI admits guardrails can fail during extended conversations and will route sensitive chats to stronger models.
- Routing to a reasoning model aims to detect distress and apply stricter response constraints.
Conspiracy Case Where ChatGPT 'Played Along'
- Jaeden describes a Wall Street Journal case where ChatGPT validated a man's conspiracy delusions, preceding a murder-suicide.
- He uses this to illustrate how typical LLMs can reinforce a user's harmful beliefs.