Multiple LLM Platforms Expose Thousands of Conversations via Share Links

# Multiple LLM Platforms Expose Thousands of Conversations via Share Links

https://gcdradar.vercel.app/incidents/multiple-llm-platforms-expose-thousands-of-conversations-via · AIR-2025-0089 · closed 2025-08-14

- Lane: Breaches & attacks
- Severity: 3 (Confirmed exploitation)
- Fix status: No fix needed
- First reported: 2025-07-31
- AI role: Attack on AI
- Products: ChatGPT, Claude, Copilot, Grok

## Summary

Share features across several major AI platforms inadvertently exposed user conversations to search engines and archiving services. The indexed data leaked personal identifiers, API keys, and business details.

## What to do

No action recorded; read the sources below and the vendor's advisory.

## Latest updates

- 2025-07-31: Reported Public Exposure of Over 100,000 LLM Conversations via Share Links Indexed by Search Engines and Archived (https://incidentdatabase.ai/cite/1186)

## Sources

- 2025-07-31 AI Incident Database: Reported Public Exposure of Over 100,000 LLM Conversations via Share Links Indexed by Search Engines and Archived https://incidentdatabase.ai/cite/1186

Verify against the vendor's advisory before changing production. Data CC BY-SA 4.0, https://gcdradar.vercel.app.