Meta AI Vulnerability Allowed Unauthorized Access to Prompts

# Meta AI Vulnerability Allowed Unauthorized Access to Prompts

https://gcdradar.vercel.app/incidents/meta-ai-vulnerability-allowed-unauthorized-access-to-prompts · AIR-2024-0033 · closed 2025-01-09

- Lane: Patch now
- Severity: 2 (Disclosed vulnerability)
- Fix status: Fix status unknown
- First reported: 2024-12-26
- AI role: Attack on AI
- Products: Meta AI

## Summary

A security flaw involving guessable prompt IDs and weak authorization checks allowed unauthorized users to view other people's chats. Meta patched the vulnerability and awarded a bug bounty.

## What to do

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

## Latest updates

- 2024-12-26: Meta AI Bug in Deployed Service Reportedly Allowed Potential Access to Other Users' Prompts and Responses (https://incidentdatabase.ai/cite/1172)

## Sources

- 2024-12-26 AI Incident Database: Meta AI Bug in Deployed Service Reportedly Allowed Potential Access to Other Users' Prompts and Responses https://incidentdatabase.ai/cite/1172

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