No Priors AI The Trust Problem Behind Grok Leaks
Nov 5, 2025
Trust issues in AI take center stage as Grok leaks expose private chats. The leaks raise questions about psychological, technological, and policy implications. Thousands of personal conversations became searchable, revealing shocking user prompts. Comparisons are made to previous incidents involving OpenAI, highlighting the irony of Grok's fall from grace. Expert advice on using private modes and safeguarding personal information takes the spotlight, ensuring users can maintain their privacy in this evolving landscape.
AI Snips
Chapters
Transcript
Episode notes
Grok Chats Became Publicly Searchable
- Jaeden Schafer recounts that Grok had thousands of leaked chat conversations become searchable on Google.
- He links the issue to a share-button behavior that made private links public and indexable.
Indexing Amplifies Worst-Case Use Cases
- Schafer highlights that media amplifies the most scandalous Grok uses because those get clicks.
- The indexed chats exposed a wide range of extreme and illicit user prompts that attracted attention.
Leaked Chats Show Harmful Prompts And Outputs
- Jaeden lists examples found in the Grok dumps including requests for making fentanyl, bomb construction, and an assassination plan.
- He notes XAI's policies prohibit such content, yet users still coaxed harmful outputs.