Skip to content
Submit incident
AII-1002-11143
ReportedReported by fewer than three independent publishers, with no official finding.

Grok chatbot leaks 370,000 user conversations via search engine indexing

A Forbes investigation discovered that over 370,000 user conversations with xAI's Grok chatbot were indexed by search engines after users clicked a share button, exposing personal data and sensitive requests because the system failed to warn that links would be automatically indexed. The leaked conversations reportedly included instructions for illegal drugs and bombs, along with plans to harm individuals, contrary to xAI's own content policies.

20 August 2025 · xAI · Privacy and surveillance
20 August 2025
AII-1002-111432 sources ↓
Quick facts
Organization
xAI (subject)
Category
Privacy and surveillance
Country
United States
Source
Exame
Date
20 August 2025
Drafted by
Raidu Incident Detection

What happened

A search of Google revealed more than 370,000 conversations with Grok indexed by search engines, according to Forbes investigation reported on 20 August 2025.

When users clicked Grok's share button to generate a link for sharing conversations via email or messaging apps, the system did not warn them that these links were also being indexed by search engines including Google, Bing, and DuckDuckGo, making conversations visible to anyone searching for them.

Reported impact

The indexed conversations included personal data and sensitive requests across medicine and psychology, as well as shared image files, spreadsheets and documents.

Some conversations contained concerning content that violated xAI's own rules, including suggestions for creating drugs, malware, and bombs, as well as methods for suicide and a detailed plan to assassinate Elon Musk.

Index analysis

Unlike the similar ChatGPT incident where share functionality required user consent to make conversations visible to search engines, Grok's share feature did not inform users that conversations would be automatically published on the Grok website and indexed by search engines.

Security experts and AI researchers, including Nathan Lambert, were affected by the lack of warning about indexation, raising concerns about data privacy and security on AI platforms.

Governance controls

As of the reporting date, xAI had not publicly responded to the incident.

Classification

Organization
xAI (subject)
Category
Privacy and surveillance
Industry
Media, social platforms and marketing
Country
United States
Harm status
Actual harm
Intent
Unintentional
Timing
Post-deployment
Harm types
Public interest
Harmed
Consumers, General public
Failure modes
Data leak, Content moderation failure
AI systems
Grok

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

Exame
Depois do Meta AI e do ChatGPT, 370 mil conversas de usuários com a IA Grok são indexadas no Google | Exame
20 August 2025
SAPO Tek
Drogas, bombas e planos para matar Elon Musk. Milhares de conversas com o Grok expostas online - Tek Notícias
20 August 2025